We calculated coccidioidomycosis incidence in 2 ways: 1) as a proportion of the population at risk, and 2) as the number of cases per person-years. Because community coccidioidomycosis rates are measured by incidence proportions, we calculated inmate rates by the same measurement to enhance comparison. Because cases per person-years is the recommended measure of disease incidence in a dynamic population, and because inmates are frequently moved from one prison to another throughout a year, calculating inmate cases per person-years gave us a measure against which to check coccidioidomycosis incidence proportions. Our concern was that the coccidioidomycosis incidence proportions might overestimate coccidioidomycosis rates in this study population.

To calculate coccidioidomycosis incidence proportion in the prisons, we derived coccidioidomycosis case counts from a surveillance system implemented in California prisons in 2007. Public health nurses assigned to CDCR prisons report coccidioidomycosis cases to the CCHCS Public Health Branch. Cases must meet the National Notifiable Diseases Surveillance System case definition for coccidioidomycosis (1). We calculated the yearly rates in prisons by dividing the surveillance-derived case counts by the published mid-year inmate populations (2). We obtained city coccidioidomycosis counts from local county health departments (F. Aranki, M. MacLean, unpub. data), and county and state coccidioidomycosis counts from data published by CDPH (3). We calculated annual community rates by dividing coccidioidomycosis counts by mid-year population estimates obtained from local health departments (for cities) and from the California Department of Finance (for counties and the state of California) (4,5). Because community data contain prison counts, prisons X and Y coccidioidomycosis counts were subtracted from their respective community coccidioidomycosis counts (city, county, and state counts), and prisons X and Y population counts were subtracted from their respective community populations. We compared prisons X and Y coccidioidomycosis incidence proportions to the incidence proportions of their surrounding communities and to those of Kern County and the state of California. Kern County coccidioidomycosis incidence proportions are benchmarks because Kern County consistently reports the highest coccidioidomycosis incidence of any county in California.

We calculated cases per person-years for prisons X and Y based on data from a cohort of inmates who had spent >1 night in 2011 at either prison X or Y (study cohort). This cohort was subject to certain exclusions: inmates who spent time at both institutions during 2011, inmates who received a coccidioidomycosis diagnosis before 2011, and inmates who received a coccidioidomycosis diagnosis in 2011 at a prison other than X or Y. To derive the total number of person-years spent at the prisons, we summed the number of days each inmate was incarcerated at prison X or Y and divided that sum by 365.

We defined primary coccidioidomycosis as an illness compatible with coccidioidomycosis that caused an inmate to seek medical attention and that was confirmed as coccidioidomycosis by a laboratory test. We collected primary cases from the CCHCS coccidioidomycosis surveillance system. We performed a cohort study to determine risk factors for primary coccidioidomycosis based on race/ethnicity, age, and whether the inmate had diabetes mellitus (DM); the latter was included in the model because studies and case series have identified an association between DM and complications of coccidioidomycosis (6–9). For this analysis, we again used the study cohort dataset that we had used to determine the coccidioidomycosis cases per person-years. Race/ethnicity, birthdate, and DM status were available for each inmate in the study cohort. Race/ethnicity is recorded on an inmate’s arrival into the CDCR prison system and is chosen by the inmate from a list of 27 race/ethnicities that includes a category called “other.”

We grouped the race/ethnicity values into 5 categories: African American, Asian/Pacific Islander, Hispanic, white, and other. In the “other” category, we included all inmates who had self-identified as “other” on entry into prison, and all inmates who had self-identified as “American Indian.” We calculated age at midyear 2011 and included age as a continuous variable in our model. DM status of each inmate was determined on the basis of laboratory (hemoglobin A1C results) and pharmacy information (diabetic medication prescriptions). To determine risks for primary coccidioidomycosis, we explored interaction by using stratified analyses and then performed logistic regression on a model that contained all variables and interaction terms.

To determine risk factors for severe and disseminated disease, we performed a case-control study. We defined severe disease as a case of coccidioidomycosis that was confined to the lungs (nondisseminated), for which the patient required >10 days of hospitalization. A patient was determined to have severe coccidioidomycosis if he was in the hospital for >10 days during which all hospital discharge International Classification of Disease, Ninth Revision (ICD-9) codes indicated nondisseminated coccidioidomycosis (ICD-9 codes 114.0 or 114.4–114.9). We defined disseminated coccidioidomycosis as disease in which the patient had a discharge ICD 9 code for disseminated coccidioidomycosis (ICD-9 codes of 114.1–114.3) for any hospitalization. Cases were derived from the CCHCS hospitalization discharge dataset for the period July 1, 2010, through April 11, 2013. This dataset contained hospitalization data for CDCR inmates incarcerated in any of California’s 33 adult prisons. Controls were patients in whom coccidioidomycosis was diagnosed in 2011 at prison X or Y who had not been hospitalized as of April 11, 2013. We evaluated the variables of race/ethnicity, age, and DM status by using logistic regression to predict severe and disseminated coccidioidomycosis. For the models for severe and disseminated coccidioidomycosis, our numbers were not robust enough to support models with interaction terms.

We used SAS 9.2 (SAS Institute; Cary, NC, USA) for all statistical analyses. The p value for statistical significance was set at ≤0.05.

For 2011, the coccidioidomycosis cases per 100,000 population for prisons X and Y were 6,934 and 3,799, respectively, 1–2 orders of magnitude higher than the rates at the other 6 prisons in 3 counties to which coccidioidomycosis was endemic (Table 1). These rates were also an order of magnitude higher than the 2011 rate for Kern County, and that difference was consistent over the period 2007–2012 (Figure).

In 2011, 16,560 inmates spent at >1 night in prison X or prison Y. Of these, 834 were excluded from the analysis because they had been given a diagnosis of coccidioidomycosis before 2011 or at another institution in 2011; 155 were excluded because they were incarcerated at both institutions during 2011. Of the remaining 15,571 inmates, 6,064 inmates had been incarcerated at prison X and 9,507 at prison Y. In all, 516 had a diagnosis of coccidioidomycosis in 2011, 304 from prison X and 212 from prison Y. The 6,064 inmates of prison X spent 4,037 person-years at the prison, and the 9,507 inmates of prison Y spent 5,464 person-years at the prison. The coccidioidomycosis cases per 100,000 person-years was 7,530 for prison X and 3,880 for prison Y.

Of the 15,571 inmates in the cohort, 6,558 (42%) were Hispanic, 4,380 (28%) were white, and 3,728 (24%) were African American. Asian/Pacific Islanders numbered 128 (1%), of which 36 were identified as Filipino. The remaining 777 (5%) inmates were categorized as other; of these, 183 (1% of the total cohort) were identified as American Indian. The median age of the inmates in the study cohort was 39 years (range 17–89 years; 10th percentile 24 years, 90th percentile 56 years).

In a univariable model, age was significantly associated with primary coccidioidomycosis (OR 1.009, 95% CI 1.002–1.016). Stratified analyses suggested an interaction between age and race/ethnicity in predicting primary coccidioidomycosis. We therefore created a model with variables for prison (prison X or Y), DM status, and the interaction term for age and race/ethnicity. Logistic regression on this model resulted in a significant association with primary coccidioidomycosis for incarceration at prison X (compared with incarceration at prison Y) and with days of incarceration at prison X or Y (Table 2). Also significant were African American and other race/ethnicity at >40 years of age and Hispanic race/ethnicity at >55 years of age. At age 55, African American, Hispanic, and other race/ethnicity were significantly associated with coccidioidomycosis with odds ratios of 2.5 (95% CI 1.7–3.6), 1.6 (95% CI 1.1–2.3), and 2.2 (95% CI 1.2–3.9), respectively, when compared to white persons (Table 3).

A total of 115 inmates had severe coccidioidomycosis, and 115 inmates had disseminated coccidioidomycosis (the equal numbers of severe and disseminated cases was coincidental). There were 474 prison X or Y inmates in whom coccidioidomycosis was diagnosed in 2011 who had not been hospitalized as of April 11, 2013. Logistic regression on a model containing DM status, race/ethnicity, and age resulted in a significant association between severe coccidioidomycosis and DM (OR = 3.2, CI = 1.8–5.8) (Table 4). Logistic regression on a model containing DM status, race/ethnicity, and age resulted in a significant association between disseminated coccidioidomycosis and African American race/ethnicity (OR = 1.9, CI = 1.1–3.4) (Table 5).