Socio-demographic factors influence chronic proton pump inhibitor use by a large population in the Netherlands

Authors


Dr P. D. Siersema, Department of Gastroenterology and Hepatology, University Medical Center Utrecht, Heidelberglaan 100, 3584 CX Utrecht, The Netherlands.
E-mail: p.d.siersema@umcutrecht.nl

Summary

Background  Chronic proton pump inhibitor (PPI) use is common in the Western world. Socio-economic status and socio-demographic factors have been shown to influence decisions related to prescribing of various drugs, but the influence of these factors on chronic PPI use is uncertain.

Aim  To study the influence of SES and socio-demographic factors on chronic PPI use.

Methods  Data were collected from a database of a Dutch health insurance company. Subjects having had at least one prescription for a PPI were identified and followed up for 6 months. Patients were then subdivided into chronic PPI users. Socio-demographic status was based on neighbourhood level of residence. Logistic regression was performed to determine socio-demographic factors associated with PPI use.

Results  A total of 2 001 787 insured individuals were included, 85 253 subjects were chronic users. Both low income (OR 1.55; CI 1.52–1.58) and low educational level (OR 1.33; CI 1.31–1.36) were associated with chronic PPI use. Other independent predictive variables included use of 10 or more concomitant medications (OR 5.33; CI 4.96–5.72) and the use of prokinetic drugs (OR 10.01; CI 9.22–10.88).

Conclusions  Patients of a lower socio-demographic status are more likely to use PPIs on a chronic basis. The observed gradient in PPIs use may reflect differences in health, healthcare use or healthcare supply.

Introduction

Gastro-oesophageal reflux disease (GERD) is a common condition in the Western world with 12% of the adult population suffering from reflux symptoms at least once a week.1 Proton pump inhibitors (PPIs) are the mainstay of therapy in both the acute and long-term treatment of this disorder. In addition, this medication is frequently prescribed in patients with functional dyspepsia, for the treatment and prevention of peptic ulcer disease and as prophylactic medication with other drugs, making them one of the most commonly prescribed drugs in North America and Western Europe.2–5

Chronic PPI use is mainly indicated for the treatment of severe GERD and as gastroprotection with non-steroidal anti-inflammatory drug (NSAID) use. For most other indications, PPIs should only be used for 4–8 weeks. It has, however, been demonstrated that the use of PPIs is generally much longer.6 This potentially excessive use of PPIs is associated with increased healthcare costs and, although long-term PPI use has a high margin of safety, may be associated with an increased risk of developing drug-related complications such as pneumonia7 and clostridium difficile-associated disease.8 It is therefore important to identify factors associated with chronic PPI use.

In the Netherlands, decisions related to prescribing and taking of various drugs have been shown to be influenced by the socio-demographic environment. For example, the use of adjuvant chemotherapy for elderly patients with stage III colon cancer is affected by socioeconomic status (SES), gender and comorbidity.9 Furthermore, patients of a lower SES are less likely to receive the so-called triple therapy for chronic heart failure than patients of a higher SES.10 A correlation between clinical and demographic characteristics of patients and PPI use has, to our knowledge, only been demonstrated in Canada, where it was reported that high-intensity PPI users were statistically more likely to be of a lower SES.11

Thus far, a study demonstrating a correlation between clinical and demographic factors and chronic PPI use in Europe has not been conducted. Therefore, in the current study, we investigated a possible correlation between these entities in a large patient cohort in the Netherlands.

Methods

Source of data

In January 2006, the Dutch government enacted the Health Insurance Act under which every person who legally lives or works in the Netherlands is obliged to buy individual health insurance of which the benefits are specified by law. Under this system, universal coverage is delivered entirely through private insurers. All citizens are required to purchase a basic package of essential healthcare services, along with ‘own-risk coverage’.12 The government lays out what must be included in the basic benefit package: physicians’ services, hospital care, medications, physiotherapy and dental treatment for children. To date, approximately 98.5% of eligible Dutch people have enrolled.13

For this study, data were collected from a large anonymized computerized database of one of the Dutch health insurance companies covering 2.8 million Dutch inhabitants, which is approximately 17% of the total population. The sample is a good representation of the general Dutch population regarding age, gender, income and educational level. Data obtained included information on patient characteristics, i.e. gender and age, postal code and pharmaceutical prescriptions. Each reimbursed prescription record contained information on the dispensed product according to the Anatomical Therapeutic Chemical classification:14 date of dispensation, dosage and number of doses provided. PPIs available in the Netherlands include omeprazole, lansoprazole, pantoprazole, rabeprazole and esomeprazole. Information on the indication of the prescription was not available.

PPI use

A retrospective cohort study was conducted covering the period between 1 January 2006 and 30 June 2007. The source population comprised all insured individuals of 18 years of age or older on 1 January 2006. Insured individuals of unknown gender, without a postal code or with missing values for any of the variables analysed were excluded from the analysis. Furthermore, all subjects had to be insured for the complete period.

All persons having had at least one prescription for a PPI between 1 January 2006 and 31 December 2006 were identified and classified as PPI users. The remaining insured were classified as non-PPI users. PPI users were then further categorized into chronic users and intermittent users. For this categorization, the first date in 2006 on which a patient obtained a prescription for a PPI was defined as the index date. PPI users were followed up for a period of 180 days following this index date and adherence for this period was calculated. The term ‘adherence’ is used to represent the degree of prescription-filling in a given interval.15 PPI adherence was calculated for patients with at least two prescriptions and was estimated as the medication possession ratio (MPR) by dividing the number of days’ supply of PPIs dispensed during the follow-up period by the number of days from the first dispensing to the end of the follow-up period.16 As the prescribed number of units per day was not available on the prescription records provided, the number of days’ supplied was inferred using quantity-supplied information in conjunction with standardized measures of daily dose as defined by the Dutch College of General Practitioners.17 For patients who remained on treatment at the end of the 180-day follow-up, the duration of treatment was censored at that date. Patients who switched from one type of PPI to another were considered as continuing on PPI therapy.

Adherent individuals were defined as those with an MPR of at least 80% over the follow-up period and considered to be chronic users.16 Partially adherent individuals were those having an MPR <80% and were considered to be intermittent users. Patients receiving only one prescription and intermittent users were excluded from further analysis. This also included patients who received a prescription which did cover 6 months or longer (i.e. chronically) as chronic use could not be confirmed by the reimbursement of a consecutive prescription.

Potential predictors of PPI use

To facilitate determination of predictors of chronic PPI use in the Netherlands, covariates in each group were examined. These included age, gender, urbanization, single person household and concurrent use of medication(s).11, 18 SES was measured by income, a useful measure of socioeconomic position, because it relates directly to the material conditions that may influence health. In addition, educational level has also been demonstrated to be an important health determinant as it is known to influence attitudes and knowledge towards health care.19 These two measures are usually closely related and therefore in this study we will mainly focus on income; however, the influence of educational level on chronic PPI use will also be taken into account.

The information on income, obtained through the Dutch Central Bureau of Statistics, was based on postal code, i.e. neighbourhood level, representing approximately 16 households. Postal codes were assigned to three income categories: low (1st–3rd decile), medium (4th–7th decile) and high (8th–10th decile), according to the average gross monthly income of the general Dutch population, with a minimum of 500 euros and a maximum 10 000 euros per month.20 Educational level was also based on postal code and classified into three categories: low (only primary and lower secondary), medium (higher secondary), high (higher vocational, university) as was urbanization: <1000 addresses/km2, 1000 to <1500 addresses/km2 and >1500 addresses/km2. Information on educational level was obtained through Geomarktprofiel, a large Dutch survey company.

The concurrent use of medications thought to contribute to the occurrence of upper gastrointestinal symptoms and complications was analysed for the 180 day follow-up period. These included NSAIDs, antiplatelet agents, prokinetic agents and selective serotonin reuptake inhibitors (SSRIs). In addition, the numbers of prescription medications during the follow-up period were totalled. Subjects were classified as polypharmacic users when they were on 10 or more chronic concomitant medications. Polypharmacy is considered to be a well-validated measure of comorbidity.11, 21

Statistical analysis

Chi-square testing was conducted to assess significant differences in the use of PPIs for covariates. Multivariate logistic regression was performed to identify factors that were independently associated with PPI use. Adjustments were made for gender (binary), age (categorical), urbanization (categorical), single person household (categorical) and concurrent medication use (binary), as they are known to be strongly associated with medication use.18, 22 Furthermore, four consecutive models were designed to explore the relationship between SES and PPI use. A first basic model contained only the SES variable. In the following two steps, demographic variables and the use of drugs less often coprescribed with PPIs were added. In the final step, the odds ratios were corrected for the use of gastrointestinal medication(s) and polypharmacy as a proxy for comorbidity.

The level of statistical significance was set at 0.05 and the uncertainty of the estimates was assessed by 95% confidence intervals. The analysis was performed in sas version 9.1 (SAS Institute, Cary, NC, USA).

Results

After applying the exclusion criteria, a total of 2 001 787 insured individuals were included. Of these, 236 122 (11.8%) reimbursed a PPI prescription in the period analysed, of whom 85 253 (36.1%) were chronic users, 74 493 (31.6%) intermittent users and 76 376 (32.3%) people obtained only one prescription in the 6-month follow-up after the index date (Figure 1). The distribution of the MPR in the cohort is shown in Figure 2.

Figure 1.

 Categorization of subjects into non-, intermittent- and chronic PPI users.

Figure 2.

 Distribution of the medical possession rate within the cohort.

The socioeconomic and clinical characteristics were compared between chronic PPI users and non-PPI users (Table 1). Both low income and low educational level were positively associated with PPI use (P < 0.001; Figure 3). PPI users were more likely to be female and to be of an older age (P < 0.001). Patients between the ages of 60 and 74 years were most likely to use PPIs as were patients using prokinetic drugs and NSAIDs (P < 0.001). Furthermore, polypharmacy was also a positive predictor for PPI use (P < 0.001).

Table 1.   Univariate analysis of variables associated with proton pump inhibitor (PPI) use in non-PPI users versus chronic PPI users
 No. patients (%)No. patients (%)Odds ratio (95% CI)
Non-PPI usersChronic PPI users
  1. NSAIDs, nonsteroidal anti-inflammatory drugs; H2RA, histamine-2-receptor antagonist; SSRI, selective serotonin reuptake inhibitor; TCA, tricyclic antidepressant.

  2. * Chi-squared test.

All patients1 765 66585 253 
Gender
 Male*879 308 (49.8)36 609 (42.9)1.00
 Female 886 357 (50.2)48 644 (57.1)1.32 (1.3–1.34)
Age
 18–29*361 343 (20.5)1247 (1.5)1.00
 30–44526 330 (29.8)7700 (9.0)4.24 (4.00–4.50)
 45–59 469 556 (26.6)22 895 (26.8)14.12 (13.3414.95)
 60–74277 747 (15.7)29 401 (34.5)30.66 (28.96–32.45)
 >75130 689 (7.4)24 010 (28.2)53.21 (50.25–56.34)
Income
 High*488 558 (27.7)16 642 (19.5)1.00
 Medium767 321 (43.4)32 204 (37.8)1.23 (1.21–1.26)
 low509 786 (28.9)36 407 (42.7)2.10 (2.06–2.14)
Educational level
 High*574 164 (32.5)21 363 (25.0)1.0 0
 Medium582 191 (33.0)26 830 (31.5)1.24 (1.22–1.26)
 Low609 310 (34.5)37 060 (43.5)1.64 (1.61–1.66)
Prevalence single  person household
 High*604 003 (34.2)21 439 (25.2)1.00
 Medium622 510 (35.3)29 675 (34.8)1.34 (1.32–1.37)
 Low539 152 (30.5)34 139 (40.0)1.78 (1.75–1.82)
Urbanization (adresses/km2)
 >1500*820 126 (46.4)40 882 (47.9)1.00
 1000 to <1500299 386 (17.0)14 093 (16.5)1.00 (0.98–1.02)
 <1000646 153 (36.6)30 338 (35.6)1.06 (1.04–1.08)
Polypharmacy
 <10 prescriptions*1 763 831 (99.9)83 201 (97.6)1.00
 ≥10 prescriptions1834 (0.1)2052 (2.4)23.72 (22.26–25.27)
NSAID use
 No*1 711 862 (97.0)74 321 (87.2)1.00
 Yes53 803 (3.0)10 932 (12.8)4.68 (4.58–4.78)
H2RA use
 No*1 755 925 (99.4)84 961 (99.7)1.00
 Yes9740 (0.6)292 (0.3)0.62 (0.55–0.70)
Prokinetic use
 No*1 764 114 (99.9)83 735 (98.2)1.00
 Yes1551 (0.1)1518 (1.8)20.62 (19.20–22.14)
SSRI use
 No*1 724 227 (97.7)80 228 (94.1)1.00
 Yes41 438 (2.3)5025 (5.9)2.61 (2.53–2.69)
TCA use
 No*1 757 345 (99.5)83 515 (98.2)1.00
 Yes8320 (0.5)1738 (2.0)4.40 (4.17–4.63)
Corticoid use
 No*1 755 749 (99.4)80 643 (94.6)1.00
 Yes9916 (0.6)4610 (5.4)10.12 (9.77–10.49)
Antiplatelet use
 No*1 664 641 (94.3)66 097 (77.5)1.00
 Yes101 024 (5.7)19 156 (22.5)4.78 (4.69–4.86)
Pulmonary medication
 No*1 735 911 (98.3)80 956 (95.0)1.00
 Yes29 754 (1.7)4297 (5.0)3.10 (3.00–3.20)
Figure 3.

 Chronic proton pump inhibitor use increases in decreasing levels of socioeconomic status.

Multivariable logistic regression showed that both low income (OR 1.55; CI 1.52–1.58) (Table 2) and low educational level (OR 1.33; CI 1.31–1.36) (results not shown) were associated with chronic PPI use. In addition, age independently influenced the odds of using a PPI (OR 34.10; CI 32.17–36.12). The same was true for gender with women being more likely to use PPIs than men (OR 1.13; CI 1.12–1.15). Other independent variables predictive of PPI use included the use of 10 or more concomitant medications (OR 5.33 CI 4.96–5.72), the use of prokinetic drugs (OR 10.01 CI 9.22–10.88) and NSAIDs (OR 3.42 CI 3.34–3.50). Urbanization was no longer associated with chronic PPI use.

Table 2.   Multivariate analysis of predictors of chronic proton pump inhibitor use
CovariateOdds ratio95% Confidence interval
  1. NSAID, nonsteroidal anti-inflammatory drugs; H2RA, histamine-2-receptor antagonist; SSRI, selective serotonin reuptake inhibitor; TCA, tricyclic antidepressant.

  2. * Reference group, representing the point of comparison with the other groups.

Gender
 Male*1.00 
 Female1.131.12–1.15
Age
 18–29*1.00 
 30–444.243.99–4.50
 45–59 13.1412.40–13.91
 60–7424.5923.22–26.04
 >7534.1032.17–36.14
Income
 High*1.00 
 Medium1.251.23–1.28
 Low1.551.52–1.58
Prevalence single person household
 High*1.00 
 Medium1.071.05–1.09
 Low1.161.13–1.18
Urbanization (adresses/km2)
 <1000*1.00 
 1000 to <15000.990.97–1.01
 >15001.031.01–1.04
Polypharmacy5.334.96–5.72
NSAID use3.423.34–3.50
H2RAs use0.160.14–0.18
Prokinetic use10.019.22–10.88
SSRI use2.212.14–2.28
TCA use2.372.24–2.52
Corticoid use4.434.25–4.61
Antiplatelet use1.871.83–1.90
Pulmonary medication1.581.52–1.64

PPI use and SES

The consecutive models that explored the relationship between SES (income and educational level) and PPI use indicated a difference in the likelihood of PPI use among different SES levels, with those receiving a lower income and of a lower educational level being more likely to be on PPIs (Table 3, model 1). The gradient persisted for both SES measurements when adjusting for predisposing variables such as age, gender, household type and urbanization level (Table 3, model 2). The difference became less prominent when taking into account concomitant use of nongastrointestinal medication (Table 3, model 3); however, SES still remained a predictive factor of chronic PPI use, even after adjusting for polypharmacy and concurrent use of gastrointestinal drugs (Table 3, model 4).

Table 3.   Prevalence (%) and (crude) odds ratios of chronic proton pump inhibitor use and the association with income- and educational level examined in logistic regression models
  Prevalence (%)Model 1 OR (95% CI)Model 2 OR (95% CI)Model 3 OR (95% CI)Model 4 OR (95% CI)
  1. Model 1: crude odds ratios (OR) (95% CI); model 2: OR (95% CI) adjusted for age, gender, household type and urbanization; model 3: OR (95% CI) adjusted for age, gender, household type, urbanization and concurrent use of nongastrointestinal medication; model 4: OR (95% CI) adjusted for age, gender, household type, urbanization, concurrent use of nongastrointestinal and gastrointestinal medication and polypharmacy as a proxy for comorbidity.

  2. PPI, proton pump inhibitor.

  3. * Chi-squared test.

  4.  Wald chi-squared test.

  5. ‡ Reference group, representing the point of comparison with the other groups.

Chronic PPI use
 Income (P)<0.001*<0.001†<0.001†<0.001†<0.001†
  High‡19.51.01.01.01.0
  Medium 37.81.23 (1.21–1.26)1.30 (1.27–1.32)1.27 (1.25–1.30)1.25 (1.23–1.28)
  Low 42.72.10 (2.06–2.14)1.70 (1.66–1.73)1.61 (1.58–1.65)1.55 (1.52–1.58)
 Education (P)<0.001*<0.001†<0.001†<0.001†<0.001†
  High‡25.01.01.01.01.0
  Medium31.51.24 (1.22–1.26)1.22 (1.20–1.24)1.20 (1.18–1.23)1.19 (1.17–1.21)
  Low43.51.64 (1.61–1.66)1.40 (1.37–1.43)1.36 (1.34–1.39)1.33 (1.31–1.36)

Discussion

The results of this study indicate that social differences in PPI use exist in the general Dutch population with patients of a lower SES being more likely to use PPI on a chronic basis.

A majority of PPIs in our cohort were prescribed by general practitioners (GPs). In the Netherlands, GPs are the ‘gatekeepers’ of the healthcare system, which means that patients first consult a GP before being referred to a specialist.23 As health insurance covers almost all inhabitants in the Netherlands and economic reasons are not considered to be a barrier for patients to consult a GP, differences should not exist in access to healthcare facilities among different income groups.

Nonetheless, it has been shown that patients of a lower SES are more likely to visit a GP than patients of a higher SES18 and that patients of a lower SES are more likely to receive a drug prescription, independent of health status.24 These findings may also, at least partially, explain why patients of a lower SES in our cohort were prescribed PPIs more frequently.

The results of this study are consistent with other studies demonstrating that the prevalence of prescribing medication increases with declining SES. This could be due to differences in health needs22 meaning that greater needs for services are met by greater use. However, in our analysis after adjusting for polypharmacy, as a proxy for health status, low SES still remained positively associated with PPI use indicating the possible existence of vertical inequity in the use of medical care (Table 3, model 4). Vertical inequity occurs when individuals with different levels of need do not consume appropriately dissimilar amounts of health care.25 In other words, the use of health care, in this case PPI use, is higher among subjects of a lower SES, even after standardization for measurable need differences.

We also found that PPI use was slightly more common among women than among men and that it increased with age. This is in agreement with previous studies which showed that gender differences were present for all types of medications and that the age gradient persisted for the prescription of medication.22 Different characteristics have been suggested to explain the gender difference. Physicians may be more willing to prescribe drugs to women than to men with the same level of health problems.26 Furthermore, Lin et al.27 demonstrated that, although women appear to have a generally similar prevalence and features of symptomatic GERD as men, the experienced severity of symptoms in women is significantly higher than in men, which could contribute to differential disease recognition and, as a result, increased PPI prescription.

Among the other factors strongly predictive of PPI use was polypharmacy, which is a proxy for comorbidity. A possible explanation could be that physicians may perceive patients with comorbidities to be at a higher risk of upper gastrointestinal complications and thus prescribe more protective medication to these patients. Secondly, patients with chronic illnesses have been shown to be more likely to seek medical care for functional gastrointestinal symptoms.28 PPIs are often prescribed for these symptoms even though their efficacy has been proven to be limited.29 This may also explain the strong association between the use of prokinetic drugs on the one hand and PPIs on the other hand. The former drugs are frequently prescribed in functional gastrointestinal disorders, sometimes as second-line therapy in patients not responding to PPI therapy. Similar results have been shown by Targownik et al.11

Concurrent use of medications that are known to contribute to the occurrence of upper gastrointestinal symptoms and the occurrence of complications was also predictive of PPI use. However, it should be noted that PPI use is only considered to be indicated as gastroprotective medication when NSAIDs are prescribed, either alone or in combination with corticosteroids, SSRIs and/or anticoagulant drugs.30 Unfortunately, we were unable to verify whether patients on concurrent medications were prescribed PPIs according to these guidelines. As expected, the use of histamine-2-receptor antagonists (H2RAs) was negatively associated with PPI use. H2RAs are frequently prescribed as initial treatment in suspected GERD and changed for PPIs when found to be ineffective.

A simplified model depicting theoretical causal relations between the variables analysed and chronic PPI use is shown in Figure 4. Health status is influenced by patient related factors, including age and gender and by SES. A low SES has a negative effect on a patients’ health status, indirectly increasing the number of GP visits, which may result in an increased number of PPI prescriptions. In addition, low SES directly increases the number of GP visits independent of patients’ health status, thereby increasing the number of PPI prescriptions.

Figure 4.

 Theoretical causal relationships between socio-demographic factors and chronic proton pump inhibitor use.

This study has several strengths. To our knowledge, this is the first study in Europe investigating the association of socioeconomic factors and PPI use. It is based on an administrative database covering a large proportion of the Dutch population. As it is a retrospective study, the behaviour of both physicians and patients was not affected by the fact that they knew they were being observed. Furthermore, the use of electronic claims data means that we did not have to rely on self-reporting of patients, avoiding recall bias.

However, there is also a downside to performing an analysis exclusively with administrative claims data, as over-the-counter medications cannot be accounted for. Nevertheless, in the Netherlands, PPIs are exclusively available on prescription. On the other hand, NSAIDs and aspirin are available over-the-counter and therefore concurrent use of these medications may have been underestimated. Furthermore, PPI use was based on the quantity supplied on the reimbursed prescriptions. It can only be assumed that the dispensed medication is actually ingested by a particular person. Another limitation is the limited information on general lifestyle factors. Known risk factors of reflux symptoms could not be adjusted for in the analysis. It can be argued that subjects of a lower SES are more often exposed to risk factors, such as smoking or obesity31 than subjects of a high SES and, as a result, will increasingly have symptomatic GERD and thus be prescribed PPIs more frequently. A positive association between low SES and risk of reflux symptoms, independent of these known risk factors has recently been demonstrated suggesting a possible role for other underlying mechanisms associated with a low SES.32 Finally, the organizational structure of healthcare systems differs greatly between countries; therefore, our results should not be extrapolated to countries with different types of healthcare systems.

In conclusion, these results contribute to the known knowledge on the association of PPI use with socio-demographic factors in a general population. Even in a healthcare system in which medical care is performed according to well established guidelines, social differences seem to exist in the prescription behaviour of PPIs. The observed social gradient in PPI use may reflect differences in health, healthcare use or healthcare supply by physicians. The information available in this study did not allow us to distinguish between these different mechanisms; therefore, further research is warranted.

Acknowledgements

Declaration of personal interests: None. Declaration of funding interests: This study was funded by an unrestricted grant from AstraZeneca, Zoetermeer, the Netherlands.

Ancillary