Effects of characteristics of the survey instrument on response rates to a mail survey of community hospitals.

A 4-factor, 16-cell experimental design was used to investigate the relationship between response rates of community hospitals to a survey conducted by the American Hospital Association (AHA) and 4 characteristics of the survey instrument, each varied dichotomously: the perceived length of the questionnaire, the order of questions, the orientation of the appeal made in the cover letter, and the presence or absence of a promise to share the results of the study with respondents. Response rate variations between the various cells were examined and multiple logistic regression was used to analyze the significance of the association between response rates and each of the four survey instrument variables while controlling for the effect of the others. At the same time, control was also maintained for the effects of five institutional characteristics of hospitals which a previous study had shown to have a significant relationship to response: bed size, location within or outside a standard metropolitan statistical area, AHA membership status, type of ownership, and form of control. The perceived length of the questionnaire and the order of questions were found to have a significant effect on response rates, but the orientation of the cover letter and a promise to share the results of the study with the respondents were found to be insignificant.     

Factors promoting consistent adherence to safe needle precautions among hospital workers.

OBJECTIVE: To examine organizational factors and occupational characteristics associated with adherence to occupational safety guidelines recommending never recapping needles. DESIGN: Mail surveys were conducted with healthcare workers (HCWs) and infection control professionals (ICPs). SETTING: The surveys were conducted at all non-federal general hospitals in Iowa, except one tertiary-care hospital. Survey data were linked to annual survey data of the American Hospital Association (AHA). PARTICIPANTS: HCWs were sampled from statewide rosters of physicians, nurses, and laboratory workers in Iowa. Eligible HCWs worked in a setting and position in which they were likely to routinely handle needles. ICPs at all hospitals in the state were surveyed. RESULTS: Ninety-nine ICPs responded (79% response rate). AHA data were available for all variables from 84 (85%) of the hospitals. Analyses were based on 1,454 HCWs who identified one of these hospitals as their primary hospital (70% response rate). Analyses were conducted using multiple logistic regression. Positive predictors of consistent adherence included infection control personnel hours per full-time-equivalent employee (odds ratio [OR], 1.03), frequency of standard precautions education (OR, 1.11), facilities providing personal protective equipment (OR, 1.82), facilities using needleless intravenous systems (OR, 1.42), and management support for safety (OR, 1.05). Negative predictors were use of "blood and body fluid precautions" isolation category (OR, 0.74) and increased job demands (OR, 0.90). CONCLUSION: Healthcare organizations can improve staff safety by investing wisely in educational programs regarding approaches to minimize these risks, providing protective equipment, and eliminating the use of blood and body fluid precautions as an isolation policy.     

Geriatric services offered by hospitals: predicting services by internal and external community characteristics.

This article explores the role of hospitals in providing geriatric services and identifies associated community and hospital characteristics. The sample is 4571 community hospitals responding to the 1995 American Hospital Association (AHA) Annual Survey. Most hospitals offer some geriatric services; the mean is 3.3. The dependent variable is an index of 13 geriatric services created from the AHA data. Independent variables are taken from the AHA survey, Area Resource File, and census data. Regression analysis explains fifteen percent of the variance. Hospital characteristics predicting provision of geriatric services are non-profit tax status, hospital bed size and system membership. Community characteristics predicting provision of geriatric services included higher population density, high percentage of county aged, and county nursing facility beds. Hospital characteristics yielded slightly higher predictive ability than did community characteristics. The findings suggest policies to increase hospital leadership in providing access to senior services require a multi-dimensional approach.     

Hospital financial position and the adoption of electronic health records

The objective of this study was to examine the relationship between financial position and adoption of electronic health records (EHRs) in 2442 acute care hospitals. The study was cross-sectional and utilized a general linear mixed model with the multinomial distribution specification for data analysis. We verified the results by also running a multinomial logistic regression model. To measure our variables, we used data from (1) the 2007 American Hospital Association (AHA) electronic health record implementation survey, (2) the 2006 Centers for Medicare and Medicaid Cost Reports, and (3) the 2006 AHA Annual Survey containing organizational and operational data. Our dependent variable was an ordinal variable with three levels used to indicate the extent of EHR adoption by hospitals. Our independent variables were five financial ratios: (1) net days revenue in accounts receivable, (2) total margin, (3) the equity multiplier, (4) total asset turnover, and (5) the ratio of total payroll to total expenses. For control variables, we used (1) bed size, (2) ownership type, (3) teaching affiliation, (4) system membership, (5) network participation, (6) fulltime equivalent nurses per adjusted average daily census, (7) average daily census per staffed bed, (8) Medicare patients percentage, (9) Medicaid patients percentage, (10) capitation-based reimbursement, and (11) nonconcentrated market. Only liquidity was significant and positively associated with EHR adoption. Asset turnover ratio was significant but, unexpectedly, was negatively associated with EHR adoption. However, many control variables, most notably bed size, showed significant positive associations with EHR adoption. Thus, it seems that hospitals adopt EHRs as a strategic move to better align themselves with their environment.     

Risk factors for Caesarean section in Italy: results of amulticentre study

A recent large increase in Caesarean section (CS) in Italy was the initial stimulus for a study to identify risk factors for CS and, if possible, to suggest strategies to counteract the rise. The study was conducted in three hospitals where a wide range of individual variables was collected from the clinical records and from personal interviews. Crude CS rates and odds ratios were evaluated for each single variable while logistic regression has been used to investigate possible confounding factors. The study involved 1316 consecutive deliveries. Crude CS rates were 29.4%, 15.7% and 16.1%. Variables identified as high risk factors were pre-eclampsia, previous CS, breech and other non-vertex presentations. Antenatal care under an obstetrician working in the same hospital, a low number of antenatal consultations, previous miscarriages, offer (by obstetrician) and request (by women) for CS showed significantly high odds ratios (ORs). Previous live births was strongly negatively associated with CS. No relationship between type of delivery and social status was observed while a physician factor was detected in all three hospitals where rates for different physicians ranged from 0% to 52.8%. Apart from the main medical indications for Caesarean section (previous CS, breech presentation), the results seem to indicate that individual practice style may be an important determinant of the wide variation in the rates of Caesarean delivery. While this may have been suspected before this study, these results are the first hard data to indicate that, in Italy, CS is widely performed for non-medical reasons.     

What characteristics have the elderly who contact friends at hospitals?]

PURPOSES: The purposes of this study are to examine two points; 1) What effects do regular visits to hospitals have on the existence of friend?. and 2) Do the elderly use hospitals as substitutes for other meeting places or independent of other places?. METHODS: Data concerning 2,447 older adults were obtained from a national survey for the elderly aged 60 and over, conducted in 1996. 1) To examine effects of regular visits to hospitals on friends, friendships as a dependent variable, regular visits to hospitals as a confounding variable, and some two way-interaction terms as independent variables were entered in logistic regression analysis. 2) To analyze relationships between hospitals and other meeting places, correspondence analyses and logistic regression were used. 3) Existence of friends whom the elderly contact at hospitals as a dependent variable, existence of friends whom they meet at other places as a confounding variable, and some two way-interaction terms as independent variables were entered in logistic regression analysis. 3) Relationships between clinics and other meeting places were also analyzed using correspondence analyses and logistic regression. RESULTS: 1. Although regular visits to hospitals did not have a significant direct impact on existence of friends, the respondents with low activities of daily living who regularly visits hospitals were more likely to have friends than ones did not visits hospitals. 2. The respondents who meet friends at hospitals were also more likely to have contacts with friends at community centers for the elderly or community organization. 3. The respondents who meet friends at clinics had similar characteristics to those who meet friends at hospitals. CONCLUSIONS: Regular visits to hospitals may contribute to maintain contacts with friends, especially in the elderly with low ADL. The elderly may use hospitals as meeting places with friends independent of other places. Hospitals may have similar functions to what clinics as meeting places with friends.     

The cesarean birth rate: influence of hospital teaching status.

Knowledge of how cesarean birth rates vary by hospital characteristics may aid in understanding and perhaps modifying some of the structural and process components of newborn delivery services to decrease the necessity of birth by cesarean procedure. To examine the influence of select hospital characteristics, data on hospital newborn deliveries in Illinois for 1986 among women 10-50 years of age inclusive (N = 130,249) were obtained from computerized hospital discharge abstract files. Characteristics of the hospitals were obtained from the annual American Hospital Association survey. Adjusting for mother's age at delivery; presence of pregnancy, labor, and delivery complications; expected primary payer; and size of hospital, women delivering in hospitals with teaching status were less likely (odds ratio = 0.76, p less than .001, 95 percent CL: 0.73, 0.79) to have a primary cesarean birth than women delivering in hospitals without this designation. A significantly lower cesarean birth rate in teaching hospitals was also observed in women of all age groups, in Medicaid and non-Medicaid women, and for most categories of delivery complications. These data suggest the need to identify the programmatic, technologic, and manpower functions associated with hospital teaching status that could decrease the likelihood of a primary cesarean delivery. The study also suggests that changes aimed at the manner of diagnosis, monitoring, and/or management of pregnancy/delivery complications may reduce the cesarean birth rate because of large differences in the primary cesarean birth rate found between teaching and other hospitals for most categories of newborn delivery complications.     

Smoking and Cancer Screening for Women Ages 42–75: Associations in the 1990–1994 National Health Interview Surveys

Background. Previous studies have found an association between current smoking and lower rates of mammography and Pap testing. The objective of this study was to investigate the association of smoking status with breast and cervical cancer screening across the 1990–1994 National Health Interview Surveys (NHIS). The NHIS provides a consistent set of independent and dependent variables across several surveys, with data collected by a similar in-person interview methodology from year to year. This report employs more databases than have been used in other analyses that have examined associations between smoking and cancer screening.Methods. Data were from the health promotion and cancer control supplements to the 1990–1994 NHIS. The target population was women ages 42–75. Associations were examined between smoking status (never, former, <1 pack/day, ≥1 pack/day) and three screening indicators: ever had a mammogram, mammogram in the past 2 years, and Pap test in the past 3 years. Sample sizes ranged from about 3000 to over 10,800 depending on the respective NHIS survey and dependent variable. Data analyses were conducted by bivariate and multiple logistic regression.Results. Women who smoked ≥1 pack of cigarettes per day were significantly less likely to have had mammography screening in all NHIS surveys, compared to women who never smoked. Adjusted odds ratios were 0.63 to 0.74 for ever had a mammogram, and 0.56 to 0.66 for mammography in the past 2 years. Women who smoked ≥1 pack per day also had lower Pap test rates than women who never smoked in 1992–1994 (adjusted odds ratios of 0.51–0.71). Results for lighter smokers were not as consistent. Former smokers often had significantly higher screening rates than never smokers.Conclusions. Research still needs to identify reasons for lower screening among women who smoke. Factors to explore include the social networks of smokers and broader health behavior patterns. Clinicians should consider heavier smoking as a marker for risk of not obtaining screening and make assessment of screening status a priority at each visit.     

Chronic non-specific lung disease and occupational exposures estimated by means of a job exposure matrix: the Zutphen Study

Information gathered in the Zutphen Study, the Dutch contribution to the Seven Countries Study that started in the 1960s, was used in this study. Of the 1266 men invited to take part in the 1985 survey, 939 (74%) participated. All participants were interviewed according to the BMRC chronic non-specific lung disease (CNSLD) questionnaire and medically examined for CNSLD complaints by a trained physician. The physician also filled in a questionnaire containing questions concerning previous treatments for asthma, bronchitis and emphysema. Exposures were generated by means of a job exposure matrix on the basis of the longest performed job and the gain and grouped into 12 exposure categories. A logistic regression analysis was performed using the occupational exposures as the sector of industry dependent variables in allowing for smoking habits, age and socioeconomic status. For the diagnosis by the physician and treatment for emphysema and or bronchitis, the strongest elevated odds ratios were found, indicating an adverse effect of the occupational exposures. In contrast, the variable 'ever treated for asthma' had odds ratios smaller than one with most of the exposure variables indicating a selection effect. In an analysis in which everyone who was ever treated for asthma was excluded an increase in the odds ratios compared with the first analysis was seen. The relationships between occupational exposures as generated by the job exposure matrix and CNSLD were stronger than those recently reported in the literature.     

Organizational Factors Associated with Sepsis Bundle Compliance: A Nationwide Study

Sepsis is one of the leading causes of mortality in the United States (U.S.), accounting for over 220 000 deaths annually. To address the issue of mortality, the Centers for Medicare and Medicaid Services (CMS) launched a core measure with an associated clinical bundle. This measure requires completion of time-sensitive bundled interventions (eg, fluid resuscitation, antibiotics). Completion of these interventions is known to improve the odds of survival. CMS reimbursement is not currently tied to a certain compliance level; however, it is under consideration. Hospitals’ compliance with the bundle is publicly reported as the SEP-1 measure. The objective of this study was to examine which hospital organizational factors were associated with high and low compliance with the SEP-1 measure. This study used a cross-sectional design with 2017 hospital-level data from the CMS Timely and Effective Care dataset, and the American Hospital Association (AHA)’s Annual Survey. A multivariate logistic regression with outcome variable of hospital compliance (high/low) with the SEP-1 measure (0 = less than overall average on SEP-1 score; 1 = greater than overall average) was performed. Control variables included: number of beds (<100, 100-250, >250), teaching status (teaching, nonteaching), technology level (high, low), intensive care unit (ICU) beds (none, 1-4, 5-14, 15-29, 30 or more), ownership (government, non-for-profit, for-profit), sepsis case volume (0-100, 101-200, 201 or more), nurse hours per patient day, and Magnet certification status (yes/no). The study included 2429 acute care, nonspecialty adult U.S. hospitals. Hospitals were excluded if they did not complete the AHA survey. The overall average SEP1 score was 48.9% on a 0-100 scale. Of 2429 hospitals, 49% were high compliance hospitals, and 51% were low compliance. Factors associated with increased odds of compliance included: for-profit ownership (OR = 2.48; 95% CI = 1.78-3.44; P < .01), higher sepsis case volume (OR 1.18, 95% CI 1.08-1.28, P < .01), smaller numbers (5-14) of ICU beds (OR 1.76, 95% CI 1.17-2.63, P = .01), and more nursing hours (OR 1.01, 95% CI 1.00-1.02, P < .01). Factors associated with decreased odds of compliance included: more than 250 hospital beds (OR 0.68, 95% CI 0.47-0.99, P = .02), and teaching status (OR 0.58, 95% CI 0.41-0.81, P < .01). For-profit ownership, higher sepsis case volumes, a smaller number of ICU beds, and more nursing hours per patient day were positively associated with SEP-1 bundle compliance. A higher number of total hospital beds (>250) and having a residency program (teaching hospital) were negatively associated with compliance. Sepsis is the most expensive condition treated in U.S. hospitals, and hospitals must consider how they can improve their outcomes. Two strategies hospitals could consider are restricting the size of their ICUs to allow for closer monitoring, and increasing their nurse staffing. Florida Nurses Foundation and Sigma Theta Tau International Theta Epsilon Chapter.     

Prevalence of back pain among fulltime United States workers.

A source of data on the health and working conditions of a probability sample of United States workers, the Quality of Employment Survey for 1972-3 (QES73), is investigated for the first time to determine which groups of workers are more or are less likely to report back pain. Estimated coefficients from a logistic regression are used to calculate odds ratios and confidence intervals for various groups. Few previous studies on back pain among United States workers control for as many potentially confounding variables as are considered in this study and few use data from a national probability sample of workers. The following independent variables are judged to be important positive correlates based on their estimated odds ratios and confidence intervals: farm, service, blue collar, and clerical work; low levels of schooling and income; jobs needing physical effort; age between and including 50 to 64; and smoking. Marital separation was found to be negatively correlated with back pain. Caution should be exercised in attempting to generalise these findings, since the back pain variable is based on respondents' subjective evaluations. Moreover, the variable does not distinguish between lower or upper back or neck pain, nor is information on the duration or frequency of pain available.     

Prevalence of back pain among fulltime United States workers

A source of data on the health and working conditions of a probability sample of United States workers, the Quality of Employment Survey for 1972-3 (QES73), is investigated for the first time to determine which groups of workers are more or are less likely to report back pain. Estimated coefficients from a logistic regression are used to calculate odds ratios and confidence intervals for various groups. Few previous studies on back pain among United States workers control for as many potentially confounding variables as are considered in this study and few use data from a national probability sample of workers. The following independent variables are judged to be important positive correlates based on their estimated odds ratios and confidence intervals: farm, service, blue collar, and clerical work; low levels of schooling and income; jobs needing physical effort; age between and including 50 to 64; and smoking. Marital separation was found to be negatively correlated with back pain. Caution should be exercised in attempting to generalise these findings, since the back pain variable is based on respondents' subjective evaluations. Moreover, the variable does not distinguish between lower or upper back or neck pain, nor is information on the duration or frequency of pain available.     

Reduction of surgical site infection rates associated with active surveillance.

OBJECTIVE: To evaluate whether surgical site infection (SSI) rates decrease in surgical departments as a result of performing active SSI surveillance. DESIGN: Retrospective multiple logistic regression analyses. SETTING: A group of 130 surgical departments of German hospitals participating in the Krankenhaus Infektions Surveillance System (KISS). METHODS: Data for 19 categories of operative procedures performed between January 1997 and June 2004 were included (119,114 operations). Active SSI surveillance was performed according to National Nosocomial Infections Surveillance system (NNIS) methods and definitions. Departments' SSI rates were calculated individually for each year of surveillance and for each operative procedure category, taking into account when the individual departments had begun their surveillance activities. Multiple logistic regression analyses on a single operation basis were carried out with stepwise variable selection to predict outcomes for patients with SSI. The variables included were as follows: the department's year of participation, NNIS risk index variables, patients' age and sex, and the hospitals' structural characteristics, such as yearly operation frequency, number of beds, and academic status. RESULTS: For 14 of 19 operative procedure categories analyzed, there was a tendency toward lower SSI rates that was associated with increasing duration of SSI surveillance. In multiple logistic regression analyses of pooled data for all operative procedures, the departments' participation in the surveillance system was a significant independent protective factor. Compared with the surveillance year 1, the SSI risk decreased in year 2 (odds ratio, 0.84; 95% confidence interval, 0.77-0.93) and in year 3 (odds ratio, 0.75; 95% confidence interval, 0.68-0.82), and there was no change in year 4. CONCLUSION: The SSI incidence was reduced by one quarter as a result of the surveillance-induced infection control efforts, which indicates the usefulness of a voluntary surveillance system.     

Characteristics of acute care hospitals with diversity plans and translation services

Hospitals provide diversity activities for a number of reasons. The authors examined community demand, resource availability, managed care, institutional pressure, and external orientation related variables that were associated with acute care hospital diversity plans and translation services. The authors used multiple logistic regression to analyze the data for 478 hospitals in the 2006 National Inpatient Sample (NIS) dataset that had available data on the racial and ethnic status of their discharges. We also used 2004 and 2006 American Hospital Association (AHA) data to measure the two dependent diversity variables and the other independent variables. We found that resource, managed care, and external orientation variables were associated with having a diversity plan and that resource, managed care, institutional, and external orientation variables were associated with providing translation services. The authors concluded that more evidence for diversity's impact, additional resources, and more institutional pressure may be needed to motivate more hospitals to provide diversity planning and translation services.     

An insight on the use of multiple logistic regression analysis to estimate association between risk factor and disease occurrence

Multiple logistic regression is an accepted statistical method for assessing association between an anticedant characteristic (risk factor) and a quantal outcome (probability of disease occurrence), statistically adjusting for potential confounding effects of other covariates. Yet the method has potential drawbacks which are not generally recognized. This article considers one important drawback of logistic regression. Specifically the so-called main effect logistic model assumes that the probability of developing disease is linearly and additively related to the risk factors on the logistic scale. This assumption stipulates that for each risk factor, the odds ratio is constant over all reference exposure levels, and that the odds ratio exposed to two or more factors is equal to the product of individual risk factor odds ratios. If the observed odds ratios in the data follow this pattern, the model-predicted odds ratios will be accurate, and the meaning of the odds ratio for each risk factor will be straightforward. But if the observed odds ratios deviate from the model assumption, the model will not fit the data accurately, and the model-predicted odds ratios will not reflect those in the data. Although satisfactory fit can always be achieved by adding to the model polynomial and product terms derived from the original risk factors, the odds ratios estimated by such an interaction logistic model are difficult to interpret, viz., the odds ratio for each risk factor depends not only on the reference exposure levels of that factor, but also on the exposure level in other factors.(ABSTRACT TRUNCATED AT 250 WORDS)     

The routine use of health risk appraisals: results from a national study of physician organizations

PURPOSE: To document use of health risk appraisals (HRAs) by U.S. physician organizations as part of their overall approach to health promotion and to identify associated organizational characteristics. METHODS: Telephone survey of 1590 physician organizations in the United States; surveys were conducted in organizations comprising 20 or more physicians and were conducted between September 2000 and September 2001 (70% response rate). Chi-square tests and logistic regression analysis were used to examine the association between organizational characteristics and routine administration of HRAs. RESULTS: Only 22.5% of physician organizations in the United States routinely administer HRAs. External quality incentives, information technology capabilities, and status as a medical group vs. an independent practice association are associated with greater odds of the routine use of HRA. DISCUSSION: Increased use of external quality incentives and information technology in physician organizations may be important in supporting the use of HRAs.     

The choice between different statistical approaches to risk-adjustment influenced the identification of outliers

OBJECTIVE: Many statistical approaches have been applied to compare health care providers' performance, but few studies have examined how the outlier status of providers depends on the choice between risk-adjustment techniques. STUDY DESIGN AND SETTING: We analyzed the recourse to breast-conserving surgery (BCS) for breast carcinoma across 31 hospitals of the Veneto Region (Italy). The following methods were compared: the ratio of observed to expected events (O/E), regression models with provider effects introduced as dummy variables obtained by standard or weighted effect coding, and multilevel analysis. RESULTS: The O/E method classified seven hospitals (one with high and six with low BCS rates) as outliers. The regression model with the weighted parameterization gave similar results, whereas through standard effect coding all odds ratios shifted and different outliers were identified. Multilevel analysis was quite conservative in identifying small hospitals with BCS rates lower than the regional mean. CONCLUSION: Whenever feasible, results obtained through different statistical methodologies should be compared. If providers are modeled as dummy variables obtained by effect coding, departures of the model intercept from the regional mean should be checked. The increasing use of multilevel models could entail a lower sensitivity in identifying low-quality outliers if a volume-outcome relationship exists.     

Patient satisfaction revisited: A multilevel approach

Patient satisfaction surveys are increasingly used for benchmarking purposes. In the Netherlands, the results of these surveys are reported at the univariate level without taking case mix factors into account. The first objective of the present study was to determine whether differences in patient satisfaction are attributed to the hospital, department or patient characteristics. Our second aim was to investigate which case mix variables could be taken into account when satisfaction surveys are carried out for benchmarking purposes. Patients who either were discharged from eight academic and fourteen general Dutch hospitals or visited the outpatient departments of the same hospitals in 2005 participated in cross-sectional satisfaction surveys. Satisfaction was measured on six dimensions of care and one general dimension. We used multilevel analysis to estimate the proportion of variance in satisfaction scores determined by the hospital and department levels by calculating intra-class correlation coefficients (ICCs). Hospital size, hospital type, population density and response rate are four case mix variables we investigated at the hospital level. We also measured the effects of patient characteristics (gender, age, education, health status, and mother language) on satisfaction. We found ICCs on hospital and department levels ranging from 0% to 4% for all dimensions. This means that only a minor part of the variance in patient satisfaction scores is attributed to the hospital and department levels. Although all patient characteristics had some statistically significant influence on patient satisfaction, age, health status and education appeared to be the most important determinants of patient satisfaction and could be considered for case mix correction. Gender, mother language, hospital type, hospital size, population density and response rate seemed to be less important determinants. The explained variance of the patient and hospital characteristics ranged from 3% to 5% for the different dimensions. Our conclusions are, first, that a substantial part of the variance is on the patient level, while only a minor part of the variance is at the hospital and department levels. Second, patient satisfaction outcomes in the Netherlands can be corrected by the case mix variables age, health status and education.     

Moscow Health Survey 2004 – social surveying under difficult circumstances

OBJECTIVES: The aim of this paper is to present the Moscow Health Survey 2004, which was designed to examine health inequalities in Moscow. In particular we want to discuss social survey problems, such as non-response, in Moscow and Russia. METHODS: Interviews, covering social and economic circumstances, health and social trust, of a stratified random sample of the greater Moscow population, aged 18+. Reasons for nonresponse were noted down with great care. Odds ratios (ORs) for self-rated health by gender and by six social dimensions were estimated separately for districts with low and high response rates. Bias due to non-response is discussed. RESULTS AND CONCLUSIONS: About one in two (53.1 %) of approached individuals could not be interviewed, resulting in 1190 completed interviews. Non-response in most Russian surveys, but perhaps particularly in Moscow, is large, partly due to fear of strangers and distrust of authorities. ORs for poor health vary significantly by gender, occupational class, education and economic hardship. We find no significant differences in these ORs when comparing districts with low and high response rates. Non-response may be a problem when estimating prevalence rates or population means, but much less so when estimating odds ratios in multivariate analyses.