A Randomized Controlled Trial of an Automated Telephone Intervention to Improve Blood Pressure Control

Authors


Abstract

The objective of this study was to evaluate the effectiveness of a telephonic outreach program to improve blood pressure (BP) control among patients with hypertension. The authors identified adults 18 years and older with uncontrolled BP within the previous 12 months. Patients received either an automated telephone call advising them to have a walk-in BP check (n=31,619) or usual care (n=33,154). The primary outcome was BP control at 4 weeks. Significantly more patients who received the intervention achieved BP control compared with the usual care group (32.5% vs 23.7%; P<.0001). Patients in the intervention arm with cardiovascular disease, chronic kidney disease, or diabetes mellitus achieved better BP control. Older age, female sex, and having a household income above the median were associated with BP control. When designing quality-improvement interventions to increase BP control rates, health care organizations should consider utilizing an automated telephone outreach campaign.

Hypertension is a significant public health challenge in the United States, with a prevalence of 29% among adults 18 years and older in 2007–2008.[1] The extent of this problem is likely to increase as the US population ages. Hypertension is associated with coronary artery disease, renal failure, and stroke and is a significant risk factor for diabetes-related complications. Although 82% of US adults with hypertension are aware of their condition, only 53% have their hypertension under control.[2] If blood pressure (BP) control was improved, morbidity and mortality from hypertension would decrease substantially.[2, 3]

Uncontrolled hypertension is often asymptomatic; therefore, people often do not solicit help from their health care provider to control their hypertension. Changes in care management may be effective in improving BP control. Glynn and colleagues conducted a meta-analysis of randomized controlled trials that evaluated different models of care aimed at achieving BP goals and the effectiveness of reminders on clinic attendance among patients with hypertension.[4] The most effective approach to BP control involved multiple organizational interventions; however, the positive results were not sustained over the long-term. The majority of trials in the meta-analysis that assessed appointment reminder systems (via telephone or postcard) showed an improvement in patient follow-up visits, with two studies resulting in improved BP control.[4] Although appointment reminder systems may be effective, less time- and resource-intensive methods may improve BP control rates with greater efficiency.

A practical strategy for hypertension control would facilitate patients having regular, walk-in clinic visits for BP monitoring using a simple intervention. Automated telephone outreach to individuals with hypertension may be a relatively effective approach in addressing BP control in a large population. The objective of this study was to evaluate the effectiveness of a large-scale telephonic outreach program to improve BP control among individuals with hypertension.

Methods

Setting

This study was conducted at Kaiser Permanente Southern California (KPSC), an integrated health delivery system that provides comprehensive care to more than 3.5 million members at 14 medical centers and nearly 200 medical offices. Members of KPSC are socioeconomically diverse and broadly representative of the general population of Southern California.[5] KPSC's institutional review board approved the study. Informed consent was waived.

Study Population

All adult members of KPSC with hypertension were potentially eligible for the study. Members with diagnosed hypertension were identified by inpatient and outpatient International Classification of Diseases, Ninth Revision, Clinical Modification (ICD-9-CM) codes (Table 1). The accuracy of ICD-9 coding was internally validated by The Permanente Medical Group.[6] To be included in the hypertension registry, members must be 18 years or older and meet at least one of the diagnostic criteria in Table 1. The hypertension registry also indicates whether a member has been diagnosed with cardiovascular disease, chronic kidney disease (CKD), or diabetes mellitus. We identified 66,304 hypertension registry members whose most recent BP within 12 months prior to the date of randomization was ≥140/90 mm Hg, or ≥130/80 mm Hg if the member had diabetes or CKD. Health plan members were not eligible for the study if they were on hemodialysis, in hospice or a skilled nursing facility, deceased, or did not have a BP value recorded within 12 months prior to the date of randomization. We randomized the eligible members on August 2, 2010, to a usual care arm (n=33,154) and an intervention arm (n=33,150) and subsequently excluded 1531 individuals (4.8%): 1528 did not have a valid telephone number and 3 were on a “do not call” list. The final study sample included 64,773 patients, with 31,619 members in the intervention arm and 33,154 members in the control arm.

Table 1. Diagnostic Criteria for Hypertension
At least one of the following:
  1. aOutpatient International Classification of Diseases, Ninth Revision, Clinical Modification (ICD-9-CM) diagnosis codes for hypertension: 401.xx, 402.xx, 403.xx, 405.xx, or 362.1. bInpatient ICD-9-CM diagnosis codes for hypertension: 401.xx, 402.xx, 403.xx, or 405.xx.

1. Two outpatienta visits within 365 days of each other with a code for hypertension, OR
2. One outpatient visit with a code for hypertension AND one inpatientb discharge diagnosis code for hypertension, OR
3. At least one prescription for an antihypertensive medication dispensed in the past 6 months AND one outpatient visit with a code for hypertension within 365 days of the dispense date, OR
4. One outpatient visit with a code for hypertension AND a member of one of the following populations: heart failure, coronary artery disease, diabetes, chronic kidney disease, or cerebrovascular accident (excluding subarachnoid, subdural, and cardioembolic)

Intervention

The goal of the intervention was to encourage BP monitoring and to consequently improve BP control in uncontrolled hypertensive members. Outreach occurred August 9 to 16, 2010, using an automated telephone messaging system. If the telephone call was answered by a live person or by a voicemail system, the automated message was delivered. Failed call attempts (ie, busy signal or no answer) resulted in a maximum of 2 additional call attempts on the same day. Telephone calls were made between 10 am and 8 pm. The content of the automated message was developed by the KPSC outreach team. The message included a greeting stating the call was from Kaiser Permanente, an invitation to have a BP measurement at a KPSC medical center, and the hours of operation of the medical center. The automated message was played by default in English with an option to listen to the message in Spanish. The message scripts may be viewed in the Appendix.

Outcome

The primary outcome was BP control (<130/80 mm Hg for patients with diabetes or CKD and <140/90 mm Hg for all other patients[7, 8]) measured at a KPSC medical center during the 4-week period following randomization. BP was measured according to KPSC's standard protocol: a trained medical assistant measures the member's BP using an automated sphygmomanometer and if the measurement is elevated, a second measurement is obtained.

Statistical Analysis

Summary statistics were calculated for sociodemographic characteristics (age, sex, race/ethnicity, preferred language, education, and income), chronic disease status, and systolic and diastolic BPs. Means and standard deviations (SDs) were calculated for continuous variables and percentages were calculated for categorical variables. Differences in characteristics were assessed using analysis of variance for continuous variables and chi-square tests for categorical variables. If a patient did not have a BP measurement during the study period, the baseline BP was carried forward and used as the outcome value. Subgroup analyses were conducted by chronic disease status (cardiovascular disease, diabetes, and CKD). Sensitivity analyses were conducted to exclude patients who did not have a BP measurement during the study. To examine BP control by demographic characteristics, all patients were included in a multivariable logistic regression analysis. Adjustment was made for the study arm in which patients were enrolled. All analyses were performed using SAS statistical software version 9.2 (SAS Institute Inc, Cary, NC).

Results

Demographic and clinical characteristics of the patients are provided in Table 2. The mean age among all patients was 61 years (range, 18 to 105 years), the majority of the patients were women and had at least a high school diploma, and half had an annual household income of $66,500 or less. Administrative records indicated that 10% of patients preferred communicating with their health care provider in Spanish. The mean (SD) systolic and diastolic BPs closest to the date of randomization were 147.9 (11.2) and 83.5 (11.7), respectively. There were no statistically significant differences between patients in the intervention arm compared with those in the usual care arm.

Table 2. Characteristics of the Study Participants
VariableaAll Patients (n=64,773)Intervention Group (n=31,619)Usual Care Group (n=33,154)P Valueb
  1. Abbreviations: BP, blood pressure; CKD, chronic kidney disease; CVD, cardiovascular disease. aValues are expressed as mean (standard deviation) or percentage. bParticipant characteristics were compared using analysis of variance for continuous variables and chi-square tests for categorical variables. cData are based on 2000 Census data geocoded at the block level.

Mean age, y61.4 (14.4)61.4 (14.4)61.4 (14.4).69
Age, y, %.67
18–4920.720.720.7
50–5925.125.125.1
60–6924.624.424.8
≥7029.729.829.5
Female, %54.154.154.2.67
Race/ethnicity, %.14
White40.941.340.5
Asian8.28.18.2
Black17.317.217.4
Hispanic25.225.125.2
Other/unknown8.58.28.7
Preferred spoken language, %.29
English88.188.387.9
Spanish9.59.39.6
Other2.52.42.5
Median household income, $c65,857 (28,826)65,979 (29,066)65,741 (28,596).30
History of CVD, %38.238.338.1.59
History of diabetes, %27.227.227.1.76
History of CKD, %9.59.49.6.34
Systolic BP pre-outreach147.9 (11.2)147.9 (11.1)148.0 (11.3).20
Diastolic BP pre-outreach83.5 (11.7)83.4 (11.7)83.6 (11.8).09
Systolic BP post-outreach142.2 (14.9)141.2 (15.1)143.1 (14.6)<.0001
Diastolic BP post-outreach80.8 (12.6)80.3 (12.6)81.3 (12.5)<.0001

In the intervention group, 51.2% of the outreach calls were answered by a live person, 45.9% were routed to a voicemail system, and 2.9% were unanswered. BP control in both study arms was 0% at the time of randomization. Table 3 shows the proportion of patients who achieved BP control at 4 weeks postrandomization. In the overall population, BP control was significantly higher among those in the intervention arm compared with the usual care arm (32.5% vs 23.7%; P<.0001). Study patients in the intervention arm who also had cardiovascular disease, CKD, or diabetes, achieved better BP control compared with those who did not receive an automated telephone message. The absolute difference in BP control rates ranged from a low of 7.7% in participants with diabetes to 8.8% in the general hypertensive population. Less than half of patients in each study arm (47.2% and 36.8% in the intervention and control arms, respectively) had a BP measurement recorded during the study follow-up period. In sensitivity analyses excluding patients who did not have a BP measurement postrandomization (52.8% and 63.2% of patients in the intervention and control arms, respectively), BP control was significantly higher among those in the intervention arm compared with the usual care arm (68.8% vs 64.5%; P<.0001).

Table 3. BP Control Rates in the Overall Study Population and Among Patients With a BP Measurement at 4 Weeks, by Chronic Disease Status
Overall PopulationIntervention (n=31,619)Usual Care (n=33,154)Absolute DifferenceP Value
  1. Abbreviations: BP, blood pressure; CKD, chronic kidney disease; CVD, cardiovascular disease.

General32.5%23.7%8.8%<.0001
CVD (n=24,754)36.0%27.6%8.4%<.0001
Diabetes (n=17,609)34.5%26.8%7.7%<.0001
CKD (n=6136)36.3%28.3%8.0%<.0001
Patients with BP measurement(n=14,929)(n=12,186)  
General68.8%64.5%4.3%<.0001
CVD (n=11,919)67.6%63.9%3.7%<.0001
Diabetes (n=8253)66.8%63.4%3.4%.002
CKD (n=3262)62.9%57.9%5.0%.004

After adjusting for study arm, certain demographic groups had increased BP control rates following the automated outreach campaign (Table 4). Characteristics associated with an increased odds of BP control included older age (odds ratio [OR], 1.23; P<.0001), female sex (OR, 1.14; P<.0001), and having a household income above the median (OR, 1.06; P=.002). Those with Asian, black and other/unknown race/ethnicity were less likely to have their BP controlled compared with whites.

Table 4. Odds Ratios of Blood Pressure Control Among Study Participants
CharacteristicOdds Ratio (95% Confidence Interval)P Value
Age, per 10 y1.23 (1.21–1.24)<.0001
Female sex1.14 (1.10–1.18)<.0001
Race/ethnicity
White1 (reference) 
Asian0.89 (0.83–0.95).0008
Black0.87 (0.83–0.92)<.0001
Hispanic1.00 (0.95–1.05).954
Other/unknown0.61 (0.56–0.65)<.0001
Preferred spoken language
English1 (reference) 
Spanish0.97 (0.90–1.04).356
Other0.87 (0.78–0.99).029
Median household income
>$65,8571.06 (1.02–1.10).002
Intervention group1.55 (1.50–1.60)<.0001

Discussion

Automated telephone messaging was effective in increasing BP control rates in a large, integrated health care system. Overall BP control rates at 4 weeks were 32.5% in the intervention arm and 23.7% in the usual care arm with slightly lower rates among participants with cardiovascular disease, diabetes, or CKD. These findings highlight the impact that an automated telephone intervention encouraging patients to have a clinic-based BP measurement could have on hypertension control rates.

The modest increase in overall BP control rates in our study was not surprising given that none of the target population had controlled hypertension at the beginning of the study. Derose and colleagues conducted a randomized controlled trial to evaluate the effectiveness of an automated messaging campaign to improve diabetes laboratory monitoring among resistant patients.[9] Compliance with laboratory testing increased as much as 20% among patients who received an automated telephone message and rose to 26% with the addition of automated letters. In a similar study, the paired call-letter intervention was more effective than the automated phone message alone in improving immunization coverage rates of children.[10] An automated telephone outreach campaign conducted among health plan members due for fecal occult blood testing increased colorectal screening rates by 23% compared with the intervention group.[11] Although the target population in these studies differed from the population in our study, they were conducted in health care delivery settings with similar electronic health record systems used to monitor utilization and outcomes.

Our results suggest that walk-in, clinic-based BP measurements may have a positive effect on hypertension control, particularly among individuals who lack regular monitoring. For example, clinic-based staff has the opportunity to educate and refer patients for follow-up care or counseling.[8] The statistically significant but small absolute difference between the two study groups may be due to medication treatment intensification or improved medication and lifestyle adherence as a result of the medical encounter. Furthermore, individuals who are motivated to have their BP measured in the clinic may also be more motivated to achieve BP control. Interestingly, older, higher-income women in our study were more likely to have controlled hypertension at the end of the follow-up period, which may be a result of these individuals having more time and resources to manage their BP. National-level data also show higher hypertension control rates among women, older-age persons (40 years and older), and those with higher income.[12]

Study Limitations and Strengths

There are several limitations to the current study. The results of this study may not be generalizable to health care settings without integrated electronic systems, which link the pieces of health information necessary for an automated outreach campaign. Additionally, the findings may not apply to populations with other conditions that may require invasive tests, more frequent monitoring, and/or a scheduled appointment. A longer follow-up period would be necessary to assess whether the intervention group maintained improved BP control and whether the comparison group might have received treatment intensification to close the hypertension control gap. Lastly, health plan members could have had a BP measurement between study selection and recruitment, which may have resulted in over-reporting BP control rates in one or both study groups.

Despite these limitations, our study has several strengths. First, our study sample was large and diverse. Second, we used a disease registry to identify patients diagnosed with hypertension, which would limit misclassification of a hypertension diagnosis. Third, we used clinic-based BP measurements, which would reduce the potential of recall bias that results from using self-reported BP measurement levels.

Conclusions

An automated telephone outreach program was effective in increasing BP control rates in patients with hypertension and uncontrolled BP. Additionally, the intervention has the potential to reach a large population at low marginal costs. Future studies assessing the effectiveness of an automated telephone intervention should assess the impact of antihypertensive medication intensification on BP control rates and should measure such changes over an extended period. When designing quality-improvement interventions to increase BP control rates, health care organizations should consider utilizing an automated telephone outreach campaign as part of a comprehensive chronic care management program.

Acknowledgements

We thank the entire Southern California Kaiser Permanente Regional Outreach and Care Support team; Rita Gevorkyan of Clinical Operations; Diana Moulder and Jocelyn Tran-Nguyen of Pharmacy Analytical Services; Carla Riggs of the Kaiser Permanente Notification System; and Jacqueline Porcel for programming support. This study was funded by Southern California Permanente Medical Group. Preliminary results of this study were presented at the 26th Annual Scientific Meeting and Exhibition of the American Society of Hypertension, May 2011, in New York, NY.

Disclosure

The authors report no specific funding in relation to this research and no conflicts of interest to disclose.

Appendix: Outreach Call Script

English version with Spanish prompt: “Hello. This is a Complete Care message from Kaiser Permanente for [name]. Para español, oprima el numero 2. We know how busy you are but it's important to remember to take care of your health. We would like to invite you to stop by your doctor's office for a blood pressure check. Staff will be available Monday through Friday 9 am to 12 noon and 1:30 pm to 4 pm. You don't need an appointment; you can just stop by.”

Spanish version: “Hola. Este es un mensaje de Cuidado Integral de Kaiser Permanente para [nombre]. Sabemos lo ocupado que está, pero es importante que recuerde cuidar su salud. Quisiéramos invitarle a usted que visite la oficina de su doctor para chequear su presión arterial. Puede acercarse a la oficina de su doctor de lunes a viernes de 9:00 de la mañana a 12:00 del medio día y de 1:30 pm a 4:00 pm. Usted no necesita cita previa; simplemente acérquese a la oficina.”

Ancillary