The 1970s saw the initiation and rapid expansion of lay health worker (LHW) programmes in low and middle income settings, stimulated by the primary health care approach adopted by the WHO at Alma-Ata (Walt 1990). LHW programmes also became more widespread in high income settings (Rosenthal 1998). Economic recession and political and policy changes throughout the developing world in the 1980s led to reduced investments in primary health care, including LHW programmes. By the mid-1980s the effectiveness and cost of such programmes was being questioned, particularly at a national level in developing countries, and several evaluations were conducted (Walt 1990; Frankel 1992). However, most of these were uncontrolled case studies that could not produce robust assessments of effectiveness due to selection bias and confounding.
Interest in community or lay health worker programmes in low and middle income countries became more prominent again in the 1990s prompted by the AIDS epidemic; the resurgence of other infectious diseases; and the failure of the formal health system to provide adequate care for people with chronic illnesses such as cancer, epilepsy and mental illness (Maher 1999; Hadley 2000). The growing emphasis on decentralisation and partnership with community based organisations was also important. In industrialised settings, a perceived need for mechanisms to deliver health care to minority communities and to support consumers for a wide range of health issues (Witmer 1995) also led to the growth of a wide range of LHW interventions. For example, a national survey in 1998 in the United States documented 12,500 LHWs in a large number of programmes (Rosenthal 1998).
Overall, however, this growth of interest in LHWs developed in the absence of robust evidence of their effects. Five systematic reviews have examined interventions with a lay health worker component, these focusing on breast cancer screening (Bonfill 2004); breastfeeding (Sikorski 2004); support during pregnancy (Hodnett 2004a; Hodnett 2004b); and the effectiveness of community health workers in the United States (Swider 2002). However, no reviews have attempted to examine the global evidence for the effects of LHWs, as compared to other interventions, for all conditions and types of interventions in primary and community health care. As new LHW programmes, for example in home-based care and treatment support, are developed (Foster 1996; Masood 1999; Unaids 1999), reliable reassurance should be sought that these interventions do more good than harm. Such interventions also have considerable direct and indirect costs. This systematic review therefore examines the effects of lay health worker (paid and voluntary) interventions in primary and community health care on health care behaviours, people's health and wellbeing, and their satisfaction with care.
We address the following question:
Are lay health workers effective in improving the delivery of health care and health care outcomes?
To answer this question, we attempted to address the following comparisons:
- LHW interventions compared to no intervention.
- LHW interventions involving activities not now undertaken by health professionals and delegated to LHWs compared to no intervention.
- LHW interventions involving activities not now performed adequately by health professionals and delegated to LHWs compared to no intervention.
- LHW interventions involving activities now performed by health professionals but delegated to LHWs to reduce resource consumption compared to the same activities performed by health professionals.
- The effectiveness of the lay health worker intervention related to intensity of training.
Criteria for considering studies for this review
Types of studies
Randomised controlled trials.
Types of participants
Types of health care providers:
Any lay health worker (paid or voluntary) including community health workers, village health workers, cancer supporters, birth attendants etc.
For the purposes of this review, a 'lay health worker' was defined as any health worker:
- carrying out functions related to health care delivery
- trained in some way in the context of the intervention
- having no formal professional or paraprofessional certificated or degreed tertiary education.
Interventions in which a health care function was performed as an extension to a participants' profession, where profession refers to work done for pay and which required a formal tertiary education (e.g. teachers provided health promotion in Schools - also see below).
Formally trained nurse aides, medical assistants, physician assistants, paramedical workers in emergency and fire services and other self-defined health professionals or health paraprofessionals. LHWs included in this review from industrialised or less industrialised settings may have received training recognised by a Ministry of Health or other certifying education, but this training did not form part of a tertiary education certificate or degree.
Interventions involving patient support groups only as these interventions were seen as different to LHW interventions. Another review (now at the protocol stage in the Consumers and Communication Review Group) will examine this area.
Interventions involving teachers delivering health promotion or related activities in schools as this large and important system of LHWs constitutes a unique group (teachers) and setting (schools) which would be better addressed in a separate review.
Interventions involving peer health counselling programmes in schools in which peers [pupils] teach other pupils about a health issue as part of the curriculum. Again, this constitutes a unique group and setting which would be better addressed in a separate review.
LHWs in non-primary level institutions (e.g. referral hospitals) because of the focus of the review on primary and community health care.
RCTs of interventions to train self-management tutors who were not lay persons (i.e. they were health professionals). Furthermore, RCTs simply comparing lay self-management with other forms of management (i.e. do not focus on the training of tutors etc.) were also excluded because they are concerned with the effects of empowering people to manage their own health issues rather than with the effects of interventions by LHWs who are providing services to others. Our review is primarily concerned with the latter. The former is being addressed by a review lay self-management review (now at the protocol stage in the Consumers and Communication Review Group). RCTs of interventions to train self-management tutors who were themselves lay persons were eligible for inclusion in this review.
Studies which measured only consumers' knowledge, attitudes or intentions, for example knowledge of what constitutes a 'healthy diet' or attitudes towards people with HIV/AIDS. These measures were not considered useful indicators of the effectiveness of LHW interventions.
Types of consumers:
There were no restrictions on the types of patients / recipients for whom data was extracted.
Types of interventions
Any intervention delivered by lay health workers and intended to promote health, manage illness or support people. An intervention was included if the description of the intervention was adequate to allow reviewers to establish that it was a lay health worker intervention (see definition above).
RCTs comparing one form of LHW intervention with another were not excluded from this review. However, these studies (n = 5) have not yet been analysed and will be considered in the next update of this review.
Types of outcome measures
Studies were included if they assessed any outcome measures within the following groups:
(1) Utilisation of lay health worker services
(2) Consultation processes
(3) Consumer satisfaction with care
(4) Health care behaviours such as types of care plans agreed, adherence to care plans (medication, dietary advice etc), attendance at follow-up consultations and health service utilisation
(5) Health care outcomes, as assessed by a variety of measures including physiological measures such as blood pressure or blood glucose levels, patient self-reports of symptom resolution or quality of life and patient self-esteem
(7) Social development measures such as the creation of support groups or the promotion of other community activities.
Search methods for identification of studies
The following electronic databases were searched:
MEDLINE (1966 - August 2001)
CENTRAL and specialised Cochrane Registers (EPOC and Consumers and Communication Review Groups) (to August 2001)
Science Citations (to August 2001)
Embase (1966 -August 2001)
CINAHL ( 1966 - August 2001)
AMED (1966- August 2001)
Leeds Health Education Effectiveness Database (http:// www.hubley.co.uk)
We retrieved documents that included both one or more terms relating to lay health workers and one or more terms suggesting a RCT. Search strategies were tailored to each database. The strategy for MEDLINE is presented as an example. Bibliographies of studies assessed for inclusion were also searched and all contacted authors were asked for details of additional studies.
MEDLINE search strategy
1.controlled clinical trial.pt
2.randomized controlled trial.pt
3.randomized controlled trials/
5.double blind method/
6.single blind method/
9.exp clinical trials/
10.(clin$ adj25 trial$).ti,ab.
11.((singl$ or doubl$ or trebl$ or tripl$) adj25 (blind$ or mask$)).ti,ab.
17. 7 or 16
20.18 not 19
21.17 not 20
22.community health aides/
23.home health aides/
24.exp voluntary workers/
27.(lay adj5 (worker? or visitor? or attendant? or aide? or support$ or personnel)).tw.
28.(birth adj1 attendant?).tw.
30.(train$ adj1 volunteer?).tw.
32.((health or care or healthcare) adj1 worker?).tw
33.((health or medical or care or nurs$ or psychiatric) adj1 (aide or aides)).tw.
34.((health or medical or care or nurs$ or psychiatric) adj1 attendant?).tw.
35.((nurs$ or care or home) adj1 support).tw
36.(support adj1 (program$ or service? or social)).tw.
38.21 and 37
Data collection and analysis
Selection of trials:
Two reviewers independently assessed the potential relevance of all titles and abstracts identified from the electronic searches. Full text copies of all articles that were identified as potentially relevant by either reviewer were retrieved.
As assessment of the eligibility of interventions can vary between reviewers, each full paper was assessed independently for inclusion by at least two reviewers. When reviewers disagreed the decision was referred to a third reviewer. If agreement could still not be reached, the decision was referred to the entire group to resolve by consensus. Where necessary, the study authors were asked for further information. Studies that were so flawed in their design or execution as to be unlikely to provide reliable data were excluded. Such exclusions are explained in the table of excluded studies.
Assessment of methodological quality:
Two reviewers independently assessed the quality of all eligible trials using the methodological quality criteria for RCTs listed in the Cochrane EPOC Review Group module (see ADDITIONAL INFORMATION, ASSESSMENT OF METHODOLOGICAL QUALITY under GROUP DETAILS). Studies were assessed as high quality if they reported allocation concealment, higher than 80% patient follow up and intention to treat analysis. Studies were assessed as low quality if they did not meet these criteria or if they did not report the information necessary for assessment.
Two reviewers independently extracted data from all included studies on to a standard form. The data were then checked against each other and, if necessary, reference was made to the original paper. Any outstanding discrepancies between the two data extraction sheets were discussed by the data extractors and resolved by consensus. If necessary, other members of the review team considered these discrepancies. Attempts were made to contact study authors to obtain any missing information.
Data relating to the following was extracted from all included studies:
(1) participants (LHWs and consumers). For LHWs this included information on the term/s used to describe the LHW, selection criteria, basic education, tasks performed. For consumers, this included the health problems / treatment received, their age and demographic details and their cultural background.
(2) health care setting (home, primary care facility or other); geographic setting (rural, formal urban or informal urban settlement) and country.
(3) study design and the key features of studies (e.g. whether allocation to groups was at the level of individual health care provider or village/suburb).
(4) intervention (specific training and ongoing monitoring and support [including duration, methods, who delivered by etc.] and the health care tasks performed with consumers). A full description of the intervention was also extracted.
(5) the number of LHWs that were approached, trained and followed up; the number of consumers enrolled at baseline and the number and proportion followed up.
(6) outcomes assessed and timing of outcome assessment.
(7) results (effects), organised into 7 areas (consultation processes, utilisation of lay health worker services, consumer satisfaction with care, health care behaviours, health status and well being, social development measures and cost).
(8) consumer involvement in the selection, training and management of the LHW interventions.
We grouped together studies that compared broadly similar types of interventions (n = 32), as listed below. The remaining eleven studies were very diverse and could not be usefully grouped.
(1) LHW interventions to promote the uptake of breast cancer screening compared with usual care.
(2) LHW interventions to promote breastfeeding compared with usual care.
(2.1) LHW interventions to promote breastfeeding up to 2 weeks post partum compared with usual care.
(2.2) LHW interventions to promote breastfeeding between 2 weeks and 6 months post partum compared with usual care.
(3) LHW interventions to promote immunization uptake in children and adults compared with usual care.
(4) LHW interventions to reduce morbidity and mortality from acute respiratory infections and malaria compared with usual care.
(5) LHW interventions to improve hypertension treatment compared with usual care.
(6) LHW interventions to provide support for recovering alcoholics compared with usual care or office based support.
(7) LHW interventions to provide support for mothers of sick children compared with usual care.
(8) LHW interventions to provide home aide services for the elderly compared with usual care or day care services
(9) LHW interventions to promote mother-child interaction / health promotion compared with usual care.
The comparisons made in the included studies did not allow us to address the other comparison groups listed under 'Objectives'.
Where feasible, the results of included studies were combined and an estimate of effect obtained. This was possible for the following four study subgroups: LHW interventions to promote the uptake of breast cancer screening; to promote breastfeeding; to promote immunization uptake; and to provide health care and treatment. Outcomes comparisons for LHW interventions to promote the uptake of breast cancer screening, breastfeeding and immunization are expressed as adherence to a beneficial health behaviour. Outcomes for the 'LHW interventions to reduce morbidity and mortality from acute respiratory infections and malaria' comparison are expressed as number of events (mortality and morbidity). Primary outcome measures were entered for all studies except Johnson 1993, where we selected a single relevant outcome (immunization status) from those measured. Only dichotomous outcomes were included in meta-analysis owing to the methodological complications involved in combining and interpreting studies using different continuous outcome measures and because of missing data in some studies. Differences in baseline variables were rare and not considered influential. Data were re-analysed on an intention-to-treat basis where possible.
Adjustment for clustering was made where necessary, assuming an intracluster correlation coefficient (ICC) of 0.02 which is typical of primary and community care interventions (Campbell 2000). Log relative risks and standard errors of the log relative risk were then calculated for both individual and adjusted cluster RCTs and analysed using the generic inverse variance method in Review Manager 4. Relative risks were preferred to odds ratios because event rates were often high and, in these circumstances, odds ratios can be difficult to interpret (Altman 1998). Random effects meta-analysis was preferred because the studies were heterogenous.
For the remaining five study subgroups (LHW interventions to improve hypertension treatment ; to provide support for recovering alcoholics; to provide support for mothers of sick children; to provide home aide services for the elderly; and to promote mother-child interaction / health promotion), the outcomes assessed and the settings in which the studies were conducted were very diverse. We therefore judged it inappropriate to combine the results of included studies quantitatively, as an overall estimate of effect would have little practical meaning. We therefore present a brief descriptive review of these subgroups.
The diversity of studies, the small number of studies in each subgroup and the limited intervention descriptions precluded examination of the relations between the characteristics of the interventions and their effects. The feasibility of such analysis will be re-assessed when the review is updated.
Description of studies
Electronic searching identified 8,637 titles or abstracts written both in English and foreign languages. Approximately 400 of these studies were considered potentially eligible for inclusion and full text articles were ordered. 43 studies fully met our inclusion criteria including 33 individual randomised controlled trials and 10 cluster randomised control trials (Caulfield 1998; Chongsuvivat' 1996; Duan 2000; Haider 2000; Kidane 2000; Lin 1997; Morrow 1999; Mtango 1986; Ramadas 2003; Voorhees 1996). All included studies were published in English language journals.
The majority of included studies (n = 27) compared a LHW intervention arm against a control arm receiving either no intervention or usual care but three studies (Leigh 1999; Olds 2002; Tudiver 1992) compared a LHW system of care against a professional system. A further thirteen studies examined more than one intervention arm, but in all cases a control group receiving no intervention services was included for comparison. Searching also identified five studies comparing different types of LHW intervention only and three studies that examined LHWs as part of a wider range of interventions. These studies were not included in this review because they do not allow the specific assessment of LHW success when compared against usual or professional care systems.
The included studies show considerable heterogeneity in terms of their setting; the health issues on which they focused; the aims, format and content of interventions; and the outcomes measured.
Fourteen different countries were represented in the included studies. Most trials took place in North America, 24 in the U.S.A. and four in Canada. A further four studies were based in the United Kingdom and one in Ireland. In the Southern hemisphere further studies were included from New Zealand (Bullock 1995), Australia (Heller 1995), Tanzania (Mtango 1986) and South Africa (Zwarenstein 2000) and in the Northern hemisphere from Thailand (Chongsuvivat' 1996), Mexico (Morrow 1999), Ethiopia (Kidane 2000), India (Ramadas 2003), Bangladesh (Haider 2000) and Taiwan (Lin 1997).
In 28 studies the intervention was delivered to patients based in their homes. Four interventions were based solely in a primary care facility (Barnes 1999; Caulfield 1998; Komaroff 1974; Von Korff 1998). A further nine studies involved a combination of home, primary care and community based interventions (Andersen 2000; Ireys 1996; Ireys 2001; Krieger 1999; Leigh 1999; Silver 1997; Wan 1980; Weinberger 1989; Wertz 1986). One study (Lapham 1995) took place in purpose run transitional housing facilities and one study (Voorhees 1996) was organised from and conducted using local church facilities.
Objective of the interventions
Studies were grouped into three broad categories based upon the primary health care intention of the intervention. The first category contained interventions designed to provide or to improve treatment for a health related condition (n=11). Included in this category were two studies using LHWs to detect and treat acute respiratory infections (Chongsuvivat' 1996; Mtango 1986). One study (Kidane 2000) examined the use of similar LHW services to detect and treat malaria, including the monitoring and provision of chemotherapy. Three studies focused on hypertension control through approaches that included physician assistant care (Komaroff 1974); screening and referral for treatment (Krieger 1999); and lifestyle education (Lin 1997). Other studies in the group used LHWs to screen and refer the elderly to formal health care (Carpenter 1990); to screen for oral cancer (Ramadas 2003); to provide therapy and care for aphasic patients (Wertz 1986); to improve existing treatment procedures for osteoarthritis (Weinberger 1989); to supervise directly observed therapy for tuberculosis patients (Zwarenstein 2000).
Interventions intended to change people's health related knowledge, attitudes or behaviours constituted a second category (n=20). Within this diverse grouping were four interventions promoting breast feeding practices (Caulfield 1998; Dennis 2002; Haider 2000; Morrow 1999). Another four provided counselling to promote the uptake of breast and cervical cancer screening services (Andersen 2000; Duan 2000; Hoare 1994; Sung 1997) and three attempted to increase immunization uptake through outreach and follow up services (Barnes 1999; Johnson 1993; Krieger 2000). Six studies evaluated the impact of LHW visiting and education programmes on mother-infant interaction and health behavioural outcomes (Barth 1991; Bullock 1995; Johnson 1993; Olds 2002; Schuler 2000; Siegel 1980). Other studies in this category offered telephone support and advice for coronary heart disease patients (Heller 1995); AIDS primary prevention strategies (Tudiver 1992); self treatment education for back pain sufferers (Von Korff 1998); and smoking cessation strategies (Voorhees 1996).
The third category includes interventions in which the LHWs' primary purpose was to provide psycho-social support (n=6). However, it should be noted that, in some studies, the intervention had the subsidiary effect of changing people's behaviour. For instance two studies evaluated LHW support to recovering substance users enrolled in treatment programmes (Lapham 1995; Leigh 1999). Three studies assessed social support for mothers of children with chronic health problems (Ireys 1996; Ireys 2001; Silver 1997) and one study providing practical and emotional support for new mothers (Morrell 2000).
Multifaceted interventions were included in a fourth category if they aimed explicitly to provide more than one type of service in roughly equal measure. Four studies provided in home services for the elderly (McNeil 1995; Nielsen 1972; Wan 1980; Williams 1992); one intervention was directed at improving growth and development among children with non-organic failure to thrive (NOFTT) (Black 1995); and one used LHW home visiting to help reduce the incidence of low birth weight babies (Graham 1992).
These intervention groups were further divided for meta-analysis, as described above.
Mode of delivery
There was great variety in the mode of intervention delivery adopted in different studies. Some trials used very specific delivery techniques, often tailored to the individual recipient. Some studies used face-to-face contact together with some form of telephone campaign whereas others set up community meetings and activities, such as themed bingo nights (Andersen 2000). However, no trials examined the passive delivery of health education, neither through the media nor through other activities like pamphlet distribution. In one study LHWs mailed educational brochures (Andersen 2000) but this was done together with numerous other interventions. In two studies (Tudiver 1992; Von Korff 1998) volunteers held educational meeting sessions at which messages about self care health behaviour were delivered. In total eleven studies examined interventions that involved some form of telephone contact. In certain studies the intervention was delivered entirely by telephone (Heller 1995) but in others telephone counselling was part of a more complex intervention and in some it was compared against personal contact methods (Barnes 1999). Twenty two studies used home visits to deliver at least part of the intervention and six studies used primary care facilities as a base for face-to-face contact. Other methods of intervention delivery included community meetings and discussion sessions (n=6), video presentations (Caulfield 1998) and family orientated activity sessions (Ireys 2001). One study (Voorhees 1996) used church meetings and facilities to deliver coronary heart disease education and a smoking cessation strategy respectively. Another (Duan 2000) used church volunteers to contact parishioners with an intervention promoting breastfeeding.
A few trial reports gave information about the organisational base of the intervention. In fourteen studies a non-governmental organisation, charitable organisation or community group was responsible for elements of the intervention including organisation, training and delivery. Consumer involvement was integral to the intervention in eleven studies (Duan 2000; Heller 1995; Ireys 1996; Ireys 2001; Kidane 2000; Lapham 1995; Silver 1997; Tudiver 1992; Von Korff 1998; Wertz 1986; Zwarenstein 2000). Most often, former sufferers of a particular health condition were recruited to deliver the intervention to current sufferers. No study recorded that patients had been involved in the selection of LHWs. However, a number of trials recruited LHWs from participant communities, often to represent its demographic characteristics.
Lay health workers
There were considerable differences in the number of LHWs employed to deliver the interventions (range = 2 (Graham 1992) to 150 (Chongsuvivat' 1996)). It is difficult to group trends in the selection and training of lay health workers. Individuals were sometimes recruited because they were familiar with a target community or because they had experience of a particular health condition. In other studies (Carpenter 1990; McNeil 1995), volunteers were recruited but no selection process appeared to be used. For some trials selection criteria were not described.
The amount of training afforded lay health workers varied greatly between trials. In some the training consisted only of an hour long introduction to the aims and principles of the intervention (Carpenter 1990), or a 2½ hour orientation session (Dennis 2002). More complex interventions however could be preceded by longer training periods. For example, two studies reported 100 hours of training (Barth 1991; Krieger 1999), while others reported eight theoretical and practical sessions (Black 1995), eight weeks of national vocational level training (Morrell 2000) or specific health condition training lasting months (Kidane 2000; Olds 2002; Morrow 1999; Ramadas 2003; Siegel 1980; Sung 1997; Wertz 1986).
A total of 210110 individuals were targeted in the 43 studies. Nineteen studies targeted interventions towards children and five delivered interventions specifically to the elderly. A large group of studies (n=15) recruited families, or mothers, of children with a specific health problem. Included in this group were studies aimed at breastfeeding mothers (Caulfield 1998; Dennis 2002; Haider 2000; Morrow 1999) and studies that attempted to improve the mother-infant relationship for mutual benefit (Barth 1991; Johnson 1993; Olds 2002; Schuler 2000; Siegel 1980; Ireys 1996; Ireys 2001; Morrell 2000; Silver 1997; Black 1995; Graham 1992). Only one intervention was designed specifically for men. This was an education programme about high risk sexual practices delivered to homosexual men in Canada (Tudiver 1992). Nearly half the interventions focused on low income and minority populations (n=19), especially those studies set in inner city U.S.A. and in middle and low income countries.
Most studies reported multiple measures of effect and many did not specify a primary outcome. For example, many different psychometric scoring systems were used and generally different trials chose to measure different aspects of psychosocial health.
1. Morbidity and mortality
Five studies recorded morbidity and mortality data as primary outcomes in evaluating LHW interventions to improve diagnosis and treatment for acute respiratory infections (ARI) and malaria (Chongsuvivat' 1996; Kidane 2000; Mtango 1986) and screening for oral cancer (Ramadas 2003). Three of these trials used measures specific to children under the age of five while the oral cancer screening trial recorded adult mortality. In response to an education program for sufferers of back pain, Von Korff 1998 recorded pain intensity, disability and mental health indicators.
2. Treatment assessment measures
Six studies assessed the effect of LHWs on successful treatment of a particular health condition. Three of these studies (Komaroff 1974; Krieger 1999; Lin 1997) evaluated hypertension management practices, recording the successful completion of treatment and referral, patient and practitioner satisfaction as well as knowledge scores and blood pressure. A theoretically similar intervention (Weinberger 1989) aimed at osteoarthritis sufferers also measured consumer satisfaction with health care, compliance with medication and the impact of their arthritis on sufferers. Zwarenstein 2000 recorded successful treatment rates for tuberculosis patients. One study (Carpenter 1990) did not focus on a single health condition and evaluated the time elderly patients spent in health care institutions and used an 'activities of daily living' scale to assess quality of life.
3. Health service uptake
Five studies (Caulfield 1998; Dennis 2002; Haider 2000; Morrell 2000; Morrow 1999) promoted breastfeeding, evaluating initiation, duration and type of the adopted practice. One study also recorded maternal satisfaction and diarrhoea incidence in infants (Morrow 1999). Three studies (Barnes 1999; Johnson 1993; Krieger 2000) assessed the effect of a LHW intervention on immunization uptake by recording individual immunization status in the study population.
4. Behaviour change
Two studies (Heller 1995; Voorhees 1996) recorded changes in cigarette smoking behaviour, in the first study as an indicator of lifestyle practices and in the second as a direct measure of a smoking cessation strategy. Tudiver 1992 assessed the effect of an education program on AIDS risk knowledge score and sexual practices.
5. Psycho-social measurements and assessment of quality of life
Many studies used psychological tests and mental health scoring systems to assess the effects of the intervention. Measures of infant-parent well being and interaction were common to six studies (Barth 1991; Bullock 1995; Johnson 1993; Olds 2002; Schuler 2000; Siegel 1980), some of which also recorded maternal drug use, satisfaction scales and child nutritional status. Four studies (Ireys 1996; Ireys 2001; Morrow 1999; Silver 1997) combined psychological measurement with an assessment of both the perceived and the actual availability of support to mothers. Four studies providing home aide services to the elderly (McNeil 1995; Nielsen 1972; Wan 1980; Williams 1992) combined psychological health and contentment scores with formalised activities of daily living scales and measures of physical health, such as aerobic capacity and number of admissions to health care institutions.
6. Drug rehabilitation
Two studies (Lapham 1995; Leigh 1999) supported recovering substance users. Efficacy of the intervention was measured by recent drug and alcohol use as well as study attrition and treatment attendance.
7. Child physical health
Black 1995 measured infant growth and development and parent-child interaction among children with non-organic failure to thrive. Graham 1992 recorded the proportion of low birth weight babies born to high risk mothers following an intervention during pregnancy.
Included studies recorded both dichotomous and continuous outcomes. Dichotomous measures were presented in studies (n=14) examining the treatment of ARI and malaria, promoting the uptake of breast and oral cancer screening, breastfeeding and immunization. Continuous outcomes were generally reported in studies promoting parent-child health, supporting mothers of sick children, supporting substance users and providing home aide services for the elderly.
Risk of bias in included studies
Assessments of the methodological quality of included studies are shown in Additional Table 1 and Table 2. Fifteen studies were assessed as high quality, with low susceptibility to bias. The remaining 29 studies were considered low quality, meaning that bias was of greater concern. Allocation concealment was 'done' in 32 studies and in the all of the remainder it was scored 'unclear'. Loss to follow was scored 'done' in 21 studies (i.e. more than 80% of patients followed up), unclear in twelve studies and not done in ten studies. Intention to treat analysis was performed in 27 studies, in twelve the procedure was not described and in four it was 'not done'.
The grouping of studies according to methodological quality is not intended as a platform for deciding which studies should be included in the meta analysis. It is rather intended to illustrate the quality range for research on the effects of LHW interventions.
Effects of interventions
Lay health workers have been employed to deliver a very wide variety of interventions in many different health care settings. Trying to group studies by intervention type is therefore problematic and a more useful discussion can be generated by concentrating on the intended outcome or objective of each study. For the purposes of discussion and meta-analysis 32 studies have been organised into ten groups, each group containing studies that used broadly similar methods to influence a single health care outcome. Meta-analysis was performed on five of the ten groups, including a total of fifteen studies. In the others, outcomes were considered too diverse to be usefully pooled. In the majority of cases it is the primary study outcome that has been included in the analysis. The additional tables list all of the study outcomes.
The effects described in this section all favour the intervention arm of the trials, unless stated otherwise.
LHW interventions to promote breast cancer screening compared with usual care (see meta view)
Four studies (Andersen 2000; Duan 2000; Hoare 1994; Sung 1997) employed LHWs to increase the uptake of breast cancer screening services. Duan 2000 analysed separately the effect on users and non users at baseline. The pooled RR for the five comparisons was 1.05 [95% confidence intervals (CI) 0.99, 1.12], providing little evidence for a beneficial effect of the intervention [p=0.10]. Heterogeneity between study outcomes was extremely low [p=0.86; I
LHW interventions to promote breastfeeding up to 2 weeks post partum compared with usual care (see meta view)
Four studies (Caulfield 1998; Dennis 2002; Haider 2000; Morrow 1999) examined the short term post partum effects of LHW interventions to promote breastfeeding. Meta-analysis indicated that breastfeeding promotion may increase the uptake of breastfeeding practices (RR=1.69 [95% CI 0.91, 3.12] p=0.10). The heterogeneity of the studies' outcomes raises doubts about the suitability of a pooled estimate (p<0.00001; I
LHW interventions to promote breastfeeding between two weeks and six months post partum compared with usual care (see meta view)
The pooled RR for the four studies that examined the longer term influence of breastfeeding promotion interventions (Dennis 2002; Haider 2000; Morrell 2000; Morrow 1999) was 2.93 [95% CI 0.88, 9.71] p=0.08). This result should be interpreted with some caution as individual study estimates ranged from 1.06 [0.64, 1.75] (Morrell 2000) to 11.64 [7.09, 19.09] (Haider 2000). Although between study heterogeneity was substantial (p<0.00001; I
LHW interventions to promote immunisation uptake in children and adults compared with usual care (see meta view)
Three studies (Barnes 1999; Johnson 1993; Krieger 2000) provide strong evidence that LHW based promotion strategies can increase the uptake of immunisation in both adults and children (RR=1.30 [95% CI 1.14, 1.48] p=0.0001). There was little heterogeneity (p=0.95; I
LHW interventions to provide health care and treatment specific to a medical condition compared with usual care (see meta view)
Two studies (Chongsuvivat' 1996; Mtango 1986) used LHWs to diagnose and treat acute respiratory infection (ARI) in children under five years old and one study (Kidane 2000) used LHWs to treat malaria episodes in children of the same age. Two of the studies (Kidane 2000; Mtango 1986) recorded all cause mortality outcome data whereas Chongsuvivat' 1996 recorded morbidity data for ARI. The pooled RR for mortality for the two studies was 0.69 [95% CI 0.51, 0.94] (p=0.02). When morbidity data were included the pooled RR remained significant (RR=0.74 [95% CI 0.58, 0.93) p=0.01) although heterogeneity affected both the mortality RR (p=0.02; I
LHWs interventions to improve treatment of hypertension compared with usual care
Evidence from the three studies using LHWs to treat hypertension (Lin 1997; Komaroff 1974; Krieger 1999) was mixed: one study reported that LHWs could significantly lower both systolic and diastolic blood pressure (Lin 1997) but another (Komaroff 1974) found no significant differences. Krieger 1999 found that significantly more patients receiving the LHW intervention (p=0.001) completed follow-up appointments with health care providers within 90 days of referral.
LHW interventions to promote mother-infant interaction / health promotion compared with usual care
Six studies examined the effect of LHWs on mother-child health promotion (Barth 1991; Bullock 1995; Johnson 1993; Olds 2002; Schuler 2000; Siegel 1980). The results of these studies were highly variable, with many of the studies reporting multiple outcome measures. The effects could therefore not be summarised and results for individual studies are presented in Additional Table 3.
LHWs providing support for recovering alcoholics compared with office based support or usual care
The effect of LHW support on recovering alcohol users was examined in two studies (Lapham 1995; Leigh 1999). Neither found a significant effect on any of the 'alcohol use' measures reported. Nor were there any significant differences in the employment or housing status, nor social stability and the use of leisure, nor the average income of participants in the control and treatment groups. However, Leigh 1999 noted that physical health and emotional function were significantly more improved in a control arm receiving 'office-based' rather than LHW support.
LHWs providing support to mothers of sick children compared with usual care
Three studies (Ireys 1996; Ireys 2001; Silver 1997) reported maternal health outcomes following interventions to provide support for mothers of sick children. Two studies (Ireys 2001; Silver 1997) reported that maternal anxiety was lower in the intervention group but this was the only significant outcome of many reported. The same two studies also reported child mental health scores. Three scores (hostility; anxiety/depression; summary score of mental health) favoured the intervention group in one study (Ireys 2001). Other differences were not significant.
LHWs providing home aide services for the elderly compared with day care services or usual care
Four studies examined home aide services for the elderly (McNeil 1995; Nielsen 1972; Wan 1980; Williams 1992) and assessed mental and physical functioning. Multiple outcomes were reported and results were very variable. Two studies recorded general mortality in the elderly: Nielsen 1972 found no significant impacts of the intervention but Wan 1980 measured significant impacts for the LHW intervention compared to control.
Owing to the range of interventions described and outcomes measured, eleven studies could not be assigned to subgroups. The outcomes for individual studies are reported in Additional Table 4.
This review identified 43 RCTs evaluating the effects of LHW interventions in primary and community health care. The diversity of included studies limited meta-analysis to outcomes for five study subgroups (n = 15 studies in total). Of these, LHW interventions to promote immunization uptake in children and adults and to improve outcomes for ARI and malaria show promising benefits, when compared with usual care. There is also evidence that they may be effective in promoting the uptake of breastfeeding and of a small, non-significant effect of LHW interventions in promoting the uptake of breast cancer screening. The effect size for LHW interventions to promote breast cancer screening uptake was comparable to that demonstrated elsewhere (Bonfill 2004). However, recent reviews suggest that mass cancer screening may not result in survival benefit (Olsen 2004). For interventions to promote any breastfeeding, our effect estimates are a little larger than those previously reported (Sikorski 2004), probably because of differences in both the studies and outcomes included in these analyses.
The remaining subgroups (LHW interventions to improve hypertension treatment; promote parent-child interaction and health; provide support for mothers of sick children; provide support for recovering alcoholics; and provide home aide services for the elderly), including 18 studies, reported many continuous and dichotomous outcomes using a wide range of indicators and measures. These outcomes were considered too diverse to allow meaningful statistical pooling. One study examining LHWs providing home aide services for the elderly suggest that these may significantly reduce mortality (Wan 1980). This promising effect needs confirmation. For the other subgroups, it is difficult to draw any conclusions on the effectiveness of interventions.
The meta-analysis findings need to be interpreted with caution for several reasons. All subgroups included few studies and many of these had small sample sizes. Furthermore, three subgroups (LHW interventions to increase breastfeeding in the first 2 weeks post-partum and between 2 weeks and 6 months post-partum and LHW interventions to improve treatment of ARI and malaria) showed significant statistical heterogeneity.
Caution is also needed in extrapolating the meta-analysis findings to large scale programmes. LHWs in experimental studies may be more carefully selected; receive substantial training and support from highly motivated project leaders; and work with carefully selected consumer groups. Furthermore, most trials did not measure the effectiveness of LHW programmes over long periods. These factors may be important to the long term success and sustainability of large routine LHW programmes (Berman 1987; Walt 1989), although such hypotheses need further evaluation. In addition, few studies described how LHW-provided services linked with other health system components, creating difficulties or at least uncertainties for scaling up.
Most studies in this review did not compare LHW interventions with similar services delivered by professionals (substitution), but rather compared LHW interventions with 'usual care'. The few studies that compared LHW programmes with similar services delivered by professionals (e.g. Leigh 1999; Olds 2002; Tudiver 1992) reported mixed findings, with different outcomes favouring professional or LHW interventions. It is therefore possible that replacing professional care with LHWs may in some circumstances do harm and this should be considered more carefully in future studies. However, any such inferences must be viewed with caution given the diversity of the studies and the multiple outcomes reported. We would suggest that the available data allow no overall conclusions to be drawn regarding the effectiveness of LHWs in substituting for professional providers. Only four studies reported cost data (Carpenter 1990; Krieger 2000; Morrell 2000; Olds 2002). These included the running costs of the intervention (Carpenter 1990); the marginal costs per additional person immunized (Krieger 2000); cost effectiveness (Morrell 2000); and the costs of the programme (Olds 2002). We are therefore unable to draw any conclusions regarding the cost of LHW interventions compared to similar interventions delivered by health professionals.
Although participants were very varied, fewer than 50% of included studies explicitly stated that they were targeting low income or minority consumer groups. This suggests that, at least for experimental programmes, LHWs are not being used only to provide services to poorer populations. Few studies reported involving consumers in the development of the interventions; the selection of LHWs; or the support of the LHW programmes and we therefore could not assess the impacts of such involvement on intervention effectiveness.
The review identified a number of methodological problems. Firstly, there are significant difficulties in locating RCTs of LHW interventions due to the poor indexing of the term 'lay health worker' within the major health literature databases and the large number of other terms used in the literature to describe LHWs. We identified over forty such terms in the course of this review. Secondly, there is no single widely accepted definition of this cadre of health workers (Love 1997; Witmer 1995). Furthermore, applying any definition to published studies is difficult as many do not describe adequately the training and background of health workers.
Thirdly, poor descriptions of the LHW interventions, particularly with regard to the training and support, precluded assessment of the relationship between health outcomes and the type and intensity of LHW training and support (Ward 2004). Information on this relationship would help those designing LHW programs. Inadequate intervention descriptions also make difficult the development of a typology of LHW training interventions, which could be useful for grouping studies for analysis. Finally, poor intervention descriptions create difficulties in exploring the inter-relationships between different elements of these complex interventions (Campbell 2000), which may include provision of information, support and treatment.
Fourthly, many studies did not clearly specify a primary outcome creating difficulties in deciding which outcomes should be included in meta-analysis (Chan 2003). Some studies assessed large numbers of outcomes, increasing the probability of finding statistically significant differences by chance. Furthermore, the diversity of the psychometric and other outcome measures used made statistical pooling of outcome data difficult. Finally, a number of cluster randomized studies reported outcomes for individual participants without adjusting for the possible effects of clustering.
Implications for practice
LHW interventions show promising benefits in promoting the uptake of immunization in children and adults and for improving outcomes for malaria and ARI in children, when compared with usual care. Health planners could consider including LHW interventions as components of health service strategies in these areas, particularly where other effective and feasible interventions do not exist. LHWs also appear promising for breastfeeding and may also reduce mortality in the elderly. They appear to have a small, and probably not clinically relevant, effect for breast cancer screening.
For other health issues, evidence of the effectiveness of LHW interventions is so far insufficient to allow recommendations for policy and practice. There is also insufficient evidence to assess which LHW training or intervention strategies are likely to be most effective. However, LHWs are most likely to be useful as a cadre of health care providers when they have an effective health care intervention to deliver. LHWs could also potentially reduce the costs of health care if substituted for professionals, by providing care at a level closer to consumers, but evidence for this is currently lacking.
Given the growing interest in the use of LHWs for a range of health issues, for example in the delivery of home care for people with HIV/AIDS (Farmer 2001, Loewenson 2004), policy makers, funders and researchers need to be encouraged to use rigorous designs in evaluating these programs.
Implications for research
For a wide range of health issues, further rigorous research on the effectiveness of LHW interventions is needed. Greater attention needs to be paid to the quality of study designs, particularly where cluster randomisation is used. Also, investigators should specify a primary outcome; consider whether the measurement of large number of related outcomes is useful; describe more thoroughly the training and support strategies used, any co-interventions and health care organisation and system issues; and assess possible harms of the interventions.
For health issues where LHW interventions demonstrate benefits, research needs to shift to understanding which components of these often multifaceted interventions are most effective. Further research should also explore the transferability of these findings to other settings and consumer groups and the effectiveness of different approaches to the training of LHWs and the delivery of LHW-led services.
Given the wide range of health issues, consumers, settings, training strategies and delivery mechanisms for LHW interventions, there is an urgent need for the development of a coherent typology of LHW interventions that could help to guide research and practice in this field. While the RCTs included in this review cover a wide range of health issues, researchers in these fields appear to be working largely in isolation from one another, as evidenced by the failure to date to attempt to assemble the global evidence on the effectiveness of these interventions. The absence of a widely accepted definition of LHWs as well as the poor conceptualisation of the field (Love 1997) contribute to this problem. There may be potential for better sharing of knowledge across hitherto isolated health specialty areas if a coherent typology of LHW interventions was developed. Such a typology would allow LHW interventions to be conceptualised in terms of their 'lay' component rather than in terms of the specific health issues on which they focus.
Greater efforts need to be made to involve consumers in the planning and support of studies of the effectiveness of LHW programs. The effects of consumer involvement also require further research.
Economic studies should accompany trials to establish the cost-effectiveness of different LHW interventions.
Studies are needed to evaluate the effectiveness of LHWs as compared to professional health care providers in delivering interventions in the fields of health education, promotion and the management of disease.
Our thanks to the many study authors who gave us additional information regarding study designs and interventions.
Past and present staff at the editorial bases for the Cochrane Effective Practice and Organisation of Care and Consumers and Communication Review Groups also provided considerable assistance and support: Hilda Bastian, Cynthia Fraser, Sophie Hill, Laura McAuley, Jessie McGowan and Graham Mowatt. Particular thanks to our contact editor, Andy Oxman, for his ongoing support and advice; to Craig Ramsay for statistical advice; and to Marina Clarke for assisting with inclusion assessments.
Administrative support was ably provided by Sylvia Louw at the Medical Research Council, South Africa and Anna Gaze at the London School of Hygiene and Tropical Medicine. Our thanks also to Simon Goudie for assistance with graphics and editing.
We are also grateful for the comments of several consumer representatives; several anonymous peer-reviewers; and many others who shared with us their insights into lay health worker programs.
Funding for this review was provided by German Technical Development (GTZ) (95.2068.5-001.00); WHO (M12/370/1); and the European Union funded AFDOT project (ICFP500A4PRO2). Additional funding for members of the review team was provided by the Medical Research Council of South Africa. Our thanks to Andy Haines at the London School of Hygiene and Tropical Medicine for his support in obtaining funding to complete the review.
Data and analyses
- Top of page
- Authors' conclusions
- Data and analyses
- What's new
- Contributions of authors
- Declarations of interest
- Sources of support
- Index terms
Last assessed as up-to-date: 7 November 2004.
Protocol first published: Issue 1, 2003
Review first published: Issue 1, 2005
Contributions of authors
SL, JD and MZ wrote the protocol, with contributions from GA and BvW. All authors assessed studies for inclusion; participated in data extraction; and contributed to data analysis. SL and PP undertook the meta-analysis with assistance from the other authors. PP and SL drafted the study report and all authors commented on this.
Declarations of interest
Sources of support
- Medical Research Council, South Africa.
- German Technical Co-operation (GTZ), Germany.
- World Health Organisation, Switzerland.
Medical Subject Headings (MeSH)
MeSH check words
* Indicates the major publication for the study