This study examined the effect of incorporating disease prevalence into the denominator when calculating hospitalization rates for hypertension and CHF. CHF has a low prevalence in the overall population and a high hospitalization rate, and hypertension has a high prevalence in the overall population and a low hospitalization rate. After incorporating disease prevalence into the rate calculation beyond age adjusting, 31 counties were shifted to a new CHF group ranking, compared with only 14 counties that were shifted to a new hypertension group ranking.
Higher ACSC hospitalization rates indicate poor quality, uncoordinated care, or insufficient access to health care (7
). Yet, the choice of rate calculation for ACSC hospitalizations can lead to different conclusions about the effectiveness of primary care services. Crude rates measure the overall burden of hospitalizations in a population; age-adjusted rates serve as a relative index of risk that roughly adjusts for age differentials in disease risk of the population (12
). These standard rate calculations capture the whole population rather than a population that is truly at risk for hospitalizations and may underestimate the overall burden of preventable hospitalizations. With appropriate disease management and lifestyle modifications, hypertension and congestive heart failure (CHF) are chronic conditions that are largely controllable in outpatient settings. Yet, the complications of CHF are more common and evident than those for hypertension; CHF is the leading cause of hospitalization in California, especially among older adults (8
Little change occurred when accounting for hypertension prevalence in the population at risk for a hospitalization. Counties with high hospitalization rates, regardless of accounting for disease prevalence, continue to be critical areas for improving outpatient care for hypertension. These findings on hypertension refute our hypothesis that adjusting for disease prevalence would highlight areas of higher disease burden. Although overall hospitalization rates are low for hypertension, a 2010 California report showed a dramatic increase in hypertension hospitalization rates from 1999 through 2008; the largest increase occurred from 2006 through 2008 (outside this study period) (8
). Future studies should still consider disease prevalence in calculating hypertension hospitalization rates.
Hospitalizations for CHF may be more preventable. After adjusting for disease prevalence, more than half of California counties changed group rankings. Furthermore, counties that reported lower group rankings by standard calculations switched to higher group rankings when adjusting for CHF prevalence. Areas such as SPA 5-West Area in Los Angeles County, Santa Barbara County, and Santa Clara County all shifted from low to high hospitalization rates. Although these are more affluent areas with high incomes and low poverty levels, the low rate of hospitalizations and high rate when adjusted for CHF prevalence may point to a higher tendency to hospitalize people with CHF. Factors such as low disease prevalence and large number of hospital beds may also fuel these hospitalization rates and thus create a hospital supply-induced care rather than a need-induced care. Although we cannot differentiate between supply-induced and need-induced care, these rate calculations are based on a patient's zip code of residence and not on referrals into an area with better hospitals. Further research is needed to explain the higher prevalence-adjusted hospitalization rates in these areas to help reduce preventable CHF hospitalizations.
This study has several limitations. First, disease prevalence is measured from a population-based survey, whereas hospitalizations are based on an administrative census count. The underdiagnosis of chronic conditions is well established from population-based surveys (14
), and so CHIS's self-reported awareness of diagnosed hypertension and CHF may underestimate the true prevalence. The rates we calculate will therefore be somewhat inflated, but we do not expect any significant bias by county. Previous literature has indicated data limitations for hospitalization data, including poor quality control and overestimation of disease trends (16
). California's OSHPD regularly conducts a series of audits to ensure validity of hospitalization data. If data reports do not meet error tolerance levels of less than 0.1%, OSHPD sends the data back to the hospital to be corrected (19
). To address the potential overestimation by OSHPD, we limited hospital records to principal diagnosis of hypertension and CHF. Although we weighted the population-based CHIS data and examine variance of the hospitalization rates, sampling biases may limit the accuracy of the findings.
Second, hospitalization rates have been shown to vary widely by sex and race/ethnicity (20
). Because of low ACSC prevalence in some counties, it was difficult to obtain stable estimates if we stratified by more than 1 demographic characteristic. We chose to adjust by age because hospitalizations vary most widely between age groups, as evident from the age-specific rates. Third, categories for CHF generated small sample sizes that may lead to high variance when incorporating CHF disease prevalence into the denominator. This issue affects rate calculations most when we stratify by age groups at the county level. We attempted to reduce the variation by using pooled CHIS data across 3 years and imputing age-specific disease prevalence where sample sizes are too small to report.
The results from this study show that disease prevalence should be incorporated into ACSC research. Previous studies have included disease prevalence as a control variable in statistical analysis rather than incorporating disease prevalence into the hospitalization rate (23
). By incorporating prevalence into hospitalization rate calculations, we can assess the true population at risk who have the disease instead of just the population at large who reside in the area. Taking into account disease prevalence can highlight areas with higher burden of chronic conditions on the health care system (both inpatient and outpatient). Because of rising health care costs and the anticipated health care reform changes to the health care system, our current tools for evaluating health services should be assessed. Our proposed method of calculation builds on existing methods and offers a relatively simple alternative by accounting for people who are truly at risk for ACSC hospitalization. By using more accurate measures, public health officials and health care providers can develop more effective community health interventions and improve outpatient care for ACSC.