Hypertension awareness and control in an inner-city African American sample

Transcription

1 Journal of Human Hypertension (1997) 11, Stockton Press. All rights reserved /97 $12.00 Hypertension awareness and control in an inner-city African American sample VN Pavlik 1, DJ Hyman 1,2, C Vallbona 1, C Toronjo 2 and K Louis 3 1 Department of Community Medicine; 2 Department of Medicine, Baylor College of Medicine, Houston, TX 77030; and 3 Texas Southern University, College of Pharmacy, Houston, TX, USA African Americans in the US are at high risk for hyper- sive medication). Among all hypertensives, 73% were tension-related morbidity and mortality. The majority of aware, 64% were on treatment, and 28% were controlled African Americans live in central city areas, and lower to 140/90 mm Hg. Of hypertensives on treatment, 43% socioeconomic status and health care utilization patterns were controlled to 140/90 mm Hg, but 72% were con- have been hypothesized to contribute to higher trolled using the criterion of 160/95 mm Hg, and 75% blood pressure (BP) levels and poorer control of treated were controlled using a diastolic pressure 90 mm Hg hypertension in this group. In order to plan an inter- only. These results are similar to those reported for vention to improve hypertension care for inner city African Americans in the most recent US national African Americans in Houston, Texas, we conducted a health survey. Males were less likely to be aware, receiving baseline survey of residents in 12 low-income ZIP code treatment and controlled than were females. areas with a 70% African American population to Although lack of awareness was associated with less determine the level of hypertension awareness, treatment frequent BP measurement, 77% of those unaware and control, and associated sociodemographic, reported a measurement within the past 2 years. The health care utilization, and medication compliance variables. majority of aware hypertensives reported frequent physment Subjects were recruited to attend a BP measure- ician contact and high compliance with medication. We and assessment of knowledge, attitudes and conclude that intervention to improve hypertension con- behaviors through random digit phone dialing in the tar- trol in this population should focus on ensuring that get ZIP code areas. Of the 962 subjects examined, 433 health providers diagnose BP and establish treatment (45%) were hypertensive (systolic BP 140 mm Hg or goals based on the current standard of 140/90 mm Hg. diastolic pressure 90 mm Hg or taking antihyperten- Keywords: hypertension; awareness; African American Introduction environment have been hypothesized to contribute to higher blood pressure (BP) levels and poorer African Americans in the United States have a hypertension control in African Americans. 5 Lack prevalence of hypertension that is almost 60% of a regular source of primary care and non-adherhigher than in white subjects. 1 Hypertension is a ence to treatment have been associated with poor major risk factor for heart disease and stroke, and hypertension control among inner city minorities. 6 9 contributes substantially to the higher cardiovascu- In 1993 a project to improve hypertension care for lar and cerebrovascular mortality observed in inner city African Americans in Houston, Texas African Americans compared to whites in the US. 2 was undertaken with funding from the National Population surveys from the 1970s to the present Heart, Lung and Blood Institute. Houston had a have shown higher rates of undetected hypertension population of 1.7 million in 1990, of which 28% was in certain age sex groups of African Americans, African American. The city is located in Harris and either comparable or somewhat lower levels of County, and the combined population of the city treatment and control 1,3 than those observed in and surrounding suburban Harris County in 1990 whites. Efforts to further improve hypertension was 2.8 million. awareness and control among African Americans The Houston Hypertension Control Project was remain an important public health priority. planned and implemented jointly by Baylor College African Americans have lower median incomes of Medicine, Departments of Community Medicine and educational attainment than the majority group and Medicine, the Harris County Hospital District in the US. Eighty-four per cent of the US African (an agency responsible for providing health care to American population reside in a metropolitan area indigent residents of the county), and Texas Southand 57% live in inner cities. 4 The combination of ern University College of Pharmacy, a historically low socioeconomic status and the stress of the urban African American institution. The goal of the project was to improve hypertension awareness, treatment, and control in the residents of a geographic Correspondence: VN Pavlik, Department of Community Mediarea defined by eight postal ZIP code areas. This area cine, Baylor College of Medicine, Houston, TX 77030, USA Received 11 November 1996; revised 2 February 1997; accepted was selected because it has a predominantly 25 February 1997 African American, low-income population, and

2 278 was included in the catchment area of a Harris proportion based on US Census data. In order to County Hospital District community health centre ensure a large enough sample of males for assessthat could provide hypertension follow-up care to ment of hypertension awareness, males were selectively individuals without a regular health care source recruited after approximately 3700 interviews identified during screening activities. A geographi- had been completed. The RDD phone survey variables cally separate, but demographically similar area were not collected on this group of males, comprised of four contiguous residential ZIP code since random sampling assumptions did not apply areas was selected to serve as a comparison community. to their responses. In 1994, a survey was carried out to obtain baseline data on the prevalence of hypertension and factors Measures associated with awareness, treatment and con- BP was measured with mercury sphygmoman- trol in the target community and the comparison ometers by research assistants who were trained and community. In this paper we report on the level of certified using a standardized protocol. Three BP awareness and control of hypertension found in the measurements were taken 2 min apart with the sub- pooled community samples, on the relationship of ject seated using an appropriately sized cuff. The sociodemographic and health care access variables diastolic pressure was recorded as the fifth Karotkoff to awareness and control, and the implications of sound. The average of the last two BP readings was these findings for improving hypertension control in used in all statistical analyses. our target population. A questionnaire administered prior to the BP measurement elicited information on subjects Subjects and methods awareness of and knowledge related to hypertension, general and hypertension-related health care Subject recruitment utilization, and other questions on attitudes and behavior. Awareness of hypertension was assessed The target population for the baseline survey was with the question, Have you ever been told by a African American adults 18 years of age and older doctor or other health professional that you have who resided in 12 inner-city, residential ZIP code hypertension, also known as high blood pressure? areas. These ZIP code areas were estimated to have To allow for the possibility that subjects had had an a population of adult African Americans elevated reading, but had not been labelled (77% of the total population of the area). The explicitly as hypertensive, a follow-up question median 1990 household income was $ compared regarding being told their pressure was borderline to $ for the county as a whole. 10 or a little high was asked of those who said no. Residents of the target ZIP code area were Hypertension treatment was determined by asking recruited by means of random digit phone dialling respondents if they were currently taking antihypertensive (RDD) to attend a community site for a BP measurement. medication, and if not, had a physician ever This recruitment method was selected prescribed antihypertensive medication. Those who because of its low cost and the relatively high phone were not currently taking medication, but had taken coverage rates in urban areas. Previous reports had it in the past, were asked to give the reasons for discontinuation. indicated that satisfactory response rates were obtainable with RDD in minority communities. 11,12 Other information elicited from subjects during It was assumed in planning the RDD survey that the interview included frequency and site of BP 70% of respondents would agree to be examined, measurements, source of regular health care, type of and of these, 70% would keep the appointment. A health insurance, and other personal characteristics, total sample size of 2100 was desired to compare such as educational level and employment. Compliance hypertension control levels in the eight intervention with the antihypertensive medication regihypertension ZIP areas and the four comparison ZIP code areas. men was assessed with a five-item scale originally The RDD component of the survey was carried out developed by Morisky et al, 13 and subsequently by a local commercial research firm. The adult in modified by Shea et al. 14 We added a sixth item the household with the most recent birthday was the relating to cost of medications. eligible respondent. Each respondent was asked a brief series of questions covering hypertension awareness and treatment, health care utilization, Statistical analysis and socioeconomic characteristics. Those phone After verifying that there were no significant differences survey respondents who identified themselves as in demographic or hypertension control varisurvey African American at the end of the RDD survey ables in the eight intervention ZIP codes and the were invited to attend a well-known community four comparison ZIP codes, we pooled the two location (either a clinic or a community recreation samples for analysis. Because the sample consisted center) for the BP measurement and extended questionnaire of individuals recruited through both random and on hypertension knowledge, attitudes and quota sampling, we conducted unweighted analy- health behavior. They were offered compensation of ses. $10.00 for attending the BP measurement session. Logistic regression analysis was used to identify As the response rate to the RDD survey was monitored, it was noted that females were over-represented in the sample compared to their estimated variables independently associated with lack of awareness and poor hypertension control. These analyses were carried out according to the methods

3 of Hosmer and Lemeshow. 15 The variables that were Table 1 Socio-demographic characteristics of survey participants considered in these analyses were those that have (n = 962) been reported to be associated with poor control Age (in years) (mean ± s.d.) 47 ± 16 among African Americans, including male sex, low Sex (male: female) 41 : 59 educational attainment, use of the emergency room Education for non-urgent health problems, and type of health Less than high school 23% insurance. The categories of health insurance com- High school graduate 69% College graduate 9% pared were: none (individual had no form of medi- Employment status cal insurance); private (the person either purchased Employed full-time 53% insurance or was insured by his/her employer); and Employed part-time or unemployed 47% public (the person had government-funded Type of medical insurance coverage). None 16% Private 52% Unless otherwise stated, hypertensives were Public 33% defined as individuals with measured BP 140 mm Hg systolic or 90 mm Hg diastolic, or who were reportedly taking antihypertensive medication. Good BP control was defined as the proportion of tensives aware, treated and controlled in the sample individuals with average systolic BP (SBP) 140 is reflected in Table 2. Overall, 73% of hyperten- mm Hg and average diastolic BP (DBP) 90 mm Hg. sives were aware, 64% were on treatment, and 28% These definitions are consistent with those used in were controlled to 140/90 mm Hg. Of those hyperthe analysis of national health examination survey tensives on treatment, 43% were controlled. Awareness, results. 1 treatment and control were substantially better among women than men. Results The results of the logistic regression analysis of variables associated with lack of awareness are Response rate and sample characteristics shown in Table 3. Being a member of the male sex, having no BP measurement within 2 years or more, Of 3273 African Americans who completed the having full-time employment, and having public introductory RDD phone survey, 2129 (65%) agreed health insurance compared to private insurance to attend a community location for BP measurement. were directly related to lack of awareness. However, In spite of multiple reminders, and offers of trans- age, educational attainment, and use of the emergency portation, only 684 (32%) completed the examination. room for non-urgent health problems were not An additional 281 persons recruited through significantly associated with lack of awareness. the quota sampling method presented for an examination, In order to assess in the detail the extent to which for a total examination sample of 962. A lack of screening for hypertension might relate to detailed analysis of the response rate in this study lack of awareness, we examined the time since last has been published elsewhere. 16 reported BP measurement by a health professional In summary, only age, employment status (with among hypertensives and non-hypertensives (Figure unemployed persons more likely to participate), and 1). As expected, aware hypertensives reported more self-reported history of hypertension were associated frequent BP measurements than those unaware and with participation. There was no relationship persons not classified as hypertensive. Unaware between sex, number of persons in the household, hypertensives were only slightly less likely than utilization of primary care, years elapsed since last non-hypertensives to have had a BP measurement BP measurement, availability of health insurance, or within the past year, and almost equally likely to educational atttainment and participation. The have had a BP measurement within the past 2 years. response rate in the examination survey was disappointingly Thus, although less frequent measurement was asso- low, but the participants did not differ ciated with lack of awareness among persons classi- from the general community in the majority of variables fied as hypertensives, nearly 80% of those classified measured, including recency of BP measure- as unaware had been measured within the previous ment. Because of the over-representation of persons 2 years. with a history of hypertension in the examination The distribution of awareness using different BP sample, the sample cannot be used to estimate definition cutoffs is shown in Figure 2. In contrast hypertension prevalence. However, this large sam- with an unawareness level of 27% when the current ple of persons recruited from the general community criterion of 140/90 mm Hg is applied, only 10% can provide useful information regarding the treat- were unaware if a BP of 160/95 mm Hg or the taking ment and control of hypertension and the variables of antihypertensive medications is used, and associated with lack of awareness and control. when a DBP of 90 mm Hg alone is used, 13% were The sociodemographic characteristics of the 962 classified as unaware. In this sample, 5% of those persons examined are shown in Table 1. The great unaware had severe BP elevations (SBP 180 majority of participants had a high school education mm Hg or DBP 110 mm Hg). or less, and 53% were employed full-time. Sixteen Table 4 contains the odds ratios and 95% confidence per cent reported having no medical insurance of intervals of variables associated with unconper any kind. trolled hypertension (BP 140/90 mm Hg) among Of the 962 individuals examined, 433 (45%) were persons reportedly taking antihypertensive medidefined as hypertensives. The proportion of hyper- cation. Of the variables considered, only older age 279

4 280 Table 2 Hypertension awareness, treatment and control in a low-income, African-American community sample and the NHANES III National Examination Sample Houston, Texas NHANES III 1 (Arican Male (%) Female (%) Total (%) Americans (n = 189) (n = 244) (n = 433) only, %) Aware On treatment Controlled to 140/90 mm Hg On antihypertensive medications and controlled Table 3 Logistic regression analysis of variables associated with lack of awareness Dependent variable = unaware of hypertension Odds ratio P -value 95% Confidence interval Age (1 year increments) , 1.02 Male sex (referent = female sex) , 3.01 Time since last BP measurement (referent = 1 year) 2 years , years , 4.56 Emergency centre use for non-emergencies , 3.09 (referent = use of primary care) Less than high school education , 1.05 (referent = above high school education) Full-time employment , 5.37 (referent = part-time employment or unemployed) Medical insurance (referent = public health insurance) None , 1.16 Private , 0.89 Figure 2 Proportion of hypertensives aware using different BP criteria. Of the 31 persons (7% of hypertensives) with severe BP elevation (BP 180/110), 16% were unaware, and 77% of those aware were taking antihyperten- Figure 1 Time since last BP check, by awareness. sive medication. Although all of the persons with severe hypertension and unaware (n = 5) were men and having no form of medical insurance were under 65 years of age, two reported having had a BP significantly associated with inadequate hyperten- measurement within the past year. The sample of sion control. The odds ratio associated with male persons with severe hypertension is too small for sex did not achieve statistical significance in this other statistical analyses. analysis (P = 0.07), indicating that after adjustment In order to determine the extent to which treat- for other variables, the discrepancy in control ment dropouts might account for some portion of between men and women reflected in Table 4 was the uncontrolled hypertension, we analysed the lessened. characteristics of aware hypertensives who reported

5 Figure 3 BP control in persons on antihypertensive medication. As in the recent US national health survey, 1 27% of hypertensives in our study were unaware, using the level of 140/90 mm Hg to define hypertension. However, the histories of BP measurements provided by subjects indicated that 77% of those classified as unaware had received a BP measurement by a health professional within the past 2 years. Although the logistic regression analysis indicated that there was a high risk of being unaware in those who had not been measured in over 2 years, the great majority of those unaware appeared to be persons with relatively mild BP elevations who had had a recent BP measurement. Birkett et al 17 reported that the prevalence of undetected hypertension was overestimated by 350% in a Canadian sample when the classification was based on a single crosssectional BP measurement, rather than repeated ser- ial measurements. It is likely that there is a similar overestimation of the rate of undetected hyperten- sion in our African American community sample. Our data suggest that the level of undetected hypertension is partially driven by the level of BP elevation that physicians recognize as requiring action. There are no recent surveys of provider diag- nostic and treatment practices in hypertension, and the manner in which physicians currently respond to mild to moderate elevations of systolic pressure in the presence of a diastolic pressure of 90 mm Hg is not known. It is possible that many physicians do not yet classify persons with mild systolic elevations in the presence of a DBP of 90 mm Hg as hypertensive, even after repeated measurements. In screening studies in the early 1970s, young African American males were found to have the highest rate of undetected hypertension of any age sex race group. 3 In the detailed analyses of response patterns in our study, we found that 79% of males compared to 89% of females reported having a BP measurement within the past 2 years. Thus, special educational messages encouraging screening among men appear to be warranted, but must be carefully targeted to the relatively small subgroup of men who are less frequent users of health care. In the present study, male sex, but not age, was independently associated with lack of awareness in a model that controlled for time since the last BP check. Since calculation of awareness of hyperten- not having seen a physician for hypertension for 2 or more years (n = 28, or 8% of those aware). Their average age was 57 years, and 50% of them were women. The mean systolic pressure in the groups was 150 mm Hg and the mean diastolic was 85 mm Hg. None of them had severe hypertension. Only 6 (21%) of these individuals had ever taken antihypertensive medication, and of those who had, two reported discontinuing medication on the phys- ician s advice. The distribution of responses to the compliance scale in persons who reported currently taking anti- hypertensive medication is shown in Table 5. The most commonly reported compliance lapse was for- getting to take a pill, and only 32% of hypertensives on treatment gave an affirmative response to two or more of the items. The total compliance score was not related to measured BP levels in treated hyper- tensives in this sample. Discussion This study evaluated hypertension awareness, treat- ment, and control in a large sample of inner-city, low-income African Americans in the southwestern United States. Because we collected detailed infor- mation on prior BP measurement, diagnosis and treatment of hypertension, as well as data on health access and utilization in general, we were able to investigate the variables that were associated with lack of awareness and poor control in this sample. BP awareness and control in African Americans 281 Table 4 Logistic regression analysis of variables associated with uncontrolled hypertension in hypertensives on medication Dependent variable = blood pressure Odds ratio P-value 95% Confidence 140/90 mm Hg interval Age (1 year increments) , 1.06 Male sex , 2.80 (referent = female sex) Emergency center use for non-emergencies , 3.45 (referent = use of primary care) Less than high school education , 2.00 (referent = above high school education) Employed full-time , 3.45 (referent = part-time employment or unemployed) Medical insurance (referent = public health insurance) None , Private , 2.43

6 282 Table 5 Distribution of responses to compliance scale 13,14 effectiveness of existing pharmacologic agents in lowering SBP. Questions % Yes In this sample, there was no overwhelming trend toward worse hypertension control with lower Do you ever forget to take your high blood 34.1 socioeconomic status. Being fully employed and betpressure pills? Are you ever careless about taking your pills? 21.9 ter educated did not seem to confer any particular Do you ever miss taking your pills when you are 15.4 advantage. As a matter of fact, individuals who were feeling better? employed full-time were less likely to be aware and Do you ever miss taking your pills because you 10.4 controlled, although the odds ratio relating employcan t afford to buy them? ment to control on treatment did not reach statistical Do you ever miss taking your pills when you are 7.5 feeling sick? a significance. As a proxy measure for access to health Do you ever miss taking your blood pressure 26.5 care, type of health insurance was not consistently medication for any reason? associated with lack of awareness or poor control. Proportion answering yes to more than two items 32.0 The logistic regression analysis indicated that persons with no health insurance or private health a Item not included in Morisky 13 or Shea 14 versions. insurance were more likely to be aware than those with public health insurance, a finding which is difficult sion is based on the self-report that one has been to interpret. This association was only told of hypertension, it is possible that males are observed if the employment variable was included more likely to under-report their hypertensive status in the model, but an interaction term between type than women. It is also possible that males receive of health insurance and employment status was not their BP measurements in a clinical context in significant. Thus, we conclude that there is little evi- which an elevated BP is not likely to be addressed dence of a consistent health care access factor in the by the physician, such as during an acute episodic risk of being unaware of hypertension in this sam- illness. ple. In terms of control, lack of any form of health The tendency for males to be less aware of hypertension insurance was a strong risk factor for poor control. is seen in both blacks and whites in the US The data suggest, therefore, that the process of care in all of the previous national health surveys. 1 In for known hypertensives without health insurance our sample, males were consistently less likely to be is not as effective as for those with public or private measured and aware, and although the odds ratio insurance. did not reach statistical significance, once under The results of our study can only be generalized treatment, they were less likely to be controlled. to individuals in households with telephones. The However, since among aware hypertensives 84% of 1990 US Census data for the surveyed ZIP code areas men and 86% women were under treatment (either indicated a phone coverage rate of 89%. We have taking antihypertensive medication or under a doc- examined a sample of public hospital emergency tor s care for hypertension), additional research on room users to attempt to measure a segment of the the poorer BP control in men is warranted. community that was unlikely to be included. These As in other studies of hypertension treatment outcomes results will be reported in detail elsewhere. in African Americans, 1,18,19 fewer than half This study of an urban, low-income group at high of the patients under treatment achieved a BP of risk for hypertension and its complications revealed 140/90 mm Hg, whereas 75% were controlled a pattern of hypertension awareness and control using the DBP level of 90 mm Hg alone, and 72% similar to that obtained in the most recent national were controlled using the criterion of BP 160/95 health survey. When we undertook the survey, we mm Hg (Figure 3). A major reason for treatment failure expected to find significant associations between could be a lack of compliance with the phys- socioeconomic and health care access variables and ician s recommendation, including frank discontinuation hypertension awareness and control. In fact, it of physician follow-up and medication appeared that, in spite of the lower socioeconomic taking. The reported compliance with antihypertensive status of this sample, the majority of individuals had medications in our sample was very high, and contact with the health system, and hypertension only a small fraction of aware hypertensives had not awareness and control were to a large extent a func- seen a physician within the past year. It seems tion of the way health providers respond to elevated likely, therefore, that physician treatment practices BP readings. Although some targeted screening pro- and efficacy of antihypertensive medications in con- grammes may still be warranted in this community, trolling systolic pressure contribute substantially to especially in males, additional improvements in lack of control to 140/90 mm Hg. Until 1988, the population hypertension control will be achieved treatment goal for hypertensives was defined only in mainly through interventions designed to change terms of DBP, and the extent to which phys- health providers hypertension detection criteria icians have adopted the therapeutic goal of SBP and treatment practices. 140 mm Hg is unknown. The clinical trials targeting control of SBP have been conducted in the elderly, 23,24 but lack of SBP control is seen in all age References groups. Additional information is needed both on 1 Burt VL et al. Trends in the prevalence, awareness, the extent to which the lack of BP control in treated treatment and control of hypertension in the adult US hypertensives is attributable to physicians failure to population. Hypertension 1995; 26: intensify drug therapy to lower SBP and on the 2 Otten MW, Teutsch SM, Williamson DF, Marks JS. The

7 effect of known risk factors on the excess mortality of 15 Hosmer DW, Lemeshow S. Applied Logistic black adults in the United States. JAMA 1990; 263: Regression. John Wiley & Sons: New York, NY, Pavlik VN et al. Response rates in a hypertension 3 Hypertension Detection and Follow-up Program prevalence survey in African-Americans recruited Coooperative Group. Blood pressure studies in 14 through random digit phone dialing. Public Health communities. JAMA 1977; 237: Reports 1996; 111: US Department of Commerce. We the Americans: 17 Birkett NJ, Donner AP, Maynard MD. Assessing hyper- Blacks. Economics and Statistics Administration, tension control in the community: the need for follow- Bureau of the Census, September up measurements to ensure clinical relevance. CMAJ 5 James SA. Psychosocial and environmental factors in 1987; 136: Black hypertension. In: Hall WD, Saunders E, Shul- 18 Stockwell DH et al. The determinants of hypertension man, NB (eds). Hypertension in Blacks. Yearbook awareness, treatment, and control in an insured popu- Medical Publishers: Chicago, 1985, pp lation. Am J Pub Health 1994; 84: Shea S et al. Predisposing factors for severe, uncon- 19 Pavlik VN, Hyman DJ, Vallbona C. Hypertension controlled hypertension in an inner-city minority poputrol in multi-ethnic primary care clinics. J Hum Hyperlation. N Engl J Med 1992; 327: tens 1996; 10 (Suppl 3): S19 S23. 7 Bennett NM, Shea S. Hypertensive emergency: case 20 The Joint National Committee on Detection, Evalucriteria, sociodemographic profile, and previous care ation, and Treatment of High Blood Pressure. Report of 100 cases. Am J Public Health 1988; 78: of the Joint National Committee on Detection, Evalu- 8 Cummings KM et al. Determinants of drug treatment maintenance among hypertensive persons in inner city ation, and Treatment of High Blood Pressure. JAMA Detroit. Pub Health Rep 1982; 97: ; 237: Gillum RF et al. Determinants of dropout rate among 21 The Joint National Committee on Detection, Evaluhypertensive patients in an urban clinic. J Commun ation, and Treatment of High Blood Pressure. The 1980 Health 1979; 5: report of the Joint National Committee on Detection, 10 US Department of Commerce, Bureau of the Census. Evaluation, and Treatment of High Blood Pressure Census. STF3B at Arch Intern Med 1980; 140: Marin G, Vanoss B, Perez-Stable EJ. Feasibility of a 22 The Joint National Committee on Detection, Evalutelephone survey to study a minority community: His- ation and Treatment of High Blood Pressure. The 1984 panics in San Francisco. Am J Pub Health 1990; 80: Report of the Joint National Committee on Detection, Evaluation, and Treatment of High Blood Pressure. 12 Shea S, Stein AD, Lantigua R, Basch CE. Reliability of Arch Intern Med; 1984; 144: the Behavioral Risk Factor Survey in a triethnic popustroke 23 SHEP Cooperative Research Group: Prevention of lation. Am J Epidemiol 1991; 133: by antihypertensive drug treatment in older per- 13 Morisky DE, Green LW, Levine DM. Concurrent and sons with isolated systolic hypertension. JAMA 1991; predictive validity of a self-reported measure of medi- 262: cation adherence. Med Care 1986; 24: Dahlöf Bet al. Morbidity and mortality in the Swedish 14 Shea S et al. Correlates of nonadherence to hyperten- trial in old patients with hypertension (STOPsion treatment in an inner-city minority population. Hypertension). Lancet 1991; 338: Am J Pub Health 1992; 82: