Journal of Crohn's and Colitis, 2015, 77–81 doi:10.1093/ecco-jcc/jju002 Advance Access publication December 4, 2014 Original Article
Original Article
Increased Ultraviolet Light Exposure is Associated With Reduced Risk of Inpatient Surgery Among Patients With Crohn’s Disease Shail M. Govania,*, Peter D.R. Higginsa, Ryan W. Stidhama, Scott J. Montainb, Akbar K. Waljeea,c Downloaded from http://ecco-jcc.oxfordjournals.org/ by guest on November 14, 2015
Department of Internal Medicine, Division of Gastroenterology, University of Michigan, Ann Arbor, MI, USA b United States Army Research Institute of Environmental Medicine, Natick, MA, USA c Veterans Affairs Center for Clinical Management Research, Ann Arbor, MI
a
*Corresponding author: Shail M. Govani, 1500 E. Medical Center Drive, 3912 Taubman Center, SPC 5362, Ann Arbor, MI 48109, USA. Tel: +1-734-615-8468; Fax: +1-734-763-7834; Email: [email protected]
Abstract Background and aims: Due to the formation of strictures and fistulas, patients with Crohn’s disease (CD) frequently need surgery. Vitamin D has been found to play a role in the degree of inflammation. We aimed to study the effect of UV exposure on the need for inpatient surgery in patients with CD. Methods: The national inpatient sample, the largest database of hospitalizations from the USA, was used to conduct a retrospective study of patients hospitalized from 2004–2011 with CD by ICD9CM code 555. Surgery was characterized as any intestinal surgery or fistula repair. An average state UV exposure was calculated for each hospitalization. Multivariate logistic regression was used to calculate the effect of UV exposure on surgery accounting for important covariates. Results: There were 481 712 hospitalizations with a primary diagnosis of CD. Of these hospitalizations, 67 751 included a relevant surgical procedure code. Mean UV exposure was statistically lower in the group undergoing surgery (4.3 units versus 4.4 units, p = 0.001). The ratio of hospitalizations per UV exposure tertile for CD was statistically different compared with all hospitalizations (p 11. The NIS does not contain patient level zip code level, but does contain hospital zip code data for 68.6% of the patients. As there is no zip code data for 31.4% of the patients, we averaged the UV index values to obtain state level averages for all hospitalizations and integrated this data into the NIS in order to perform our main analysis. As a sensitivity analysis, we also performed multivariable modeling using 3-digit zip code data, excluding states with high standard deviations of UV index (California and Colorado), excluding perianal fistula repairs, and excluding Hawaii (because it has very few hospitalizations). As the NIS is a national de-identified dataset, IRB approval was not obtained.
2.2 Statistical analysis SAS 9.3 (Cary, NC, USA) was used to perform the statistical analysis. Complete case analysis was used for all testing. PROC SURVEYFREQ (Rao–Scott chi-square test) and SURVEYREG (t-test) were used to compare the baseline characteristics of patients with CD who underwent surgery versus those who did not. PROC SURVEYLOGISTIC was used to perform univariate and multivariable modeling to adjust for other important factors. Appropriate weighting was used to calculate national estimates with domain level analysis to ensure accurate standard error estimates.
3. Results During the study period, there were ~481 712 hospitalizations with a primary diagnosis of Crohn’s disease, with 14.1% of hospitalizations including an intestinal surgery or fistula repair. A comparison of patients with CD who underwent surgery versus those who did not is displayed in Table 1. Patients who underwent a surgery were more likely to be younger and male. There were was a trend in the income status of patients, with patients from the lowest income status less likely to undergo surgery. Unsurprisingly, patients who underwent surgery had longer lengths of stay and hospital costs. There was a statistically significant lower mean UV index exposure in patients who underwent surgery. Among patients who underwent surgery, the average UV index was 4.3 compared with an average UV index of 4.4 among those who did not receive an operation (p = 0.001).
Table 1. Comparison of patients with CD who underwent a surgery versus those who did not. Crohn’s disease visits not including a surgery (n = 413 961)
Crohn’s disease visit including a surgery (n = 67 751)
Age (years) 40.5 (±0.2) 41.1 (±0.2) Female (%) 57.5 (±0.3) 52.0 (±0.5) Race/Ethnicity (%) Caucasian 76.9 (±0.6) 84.1 (±0.9) African American 13.5 (±0.5) 8.7 (±0.4) Hispanic 5.8 (±0.4) 3.5 (±0.3) Other 3.7 (±0.2) 3.6 (±0.7) Payer information (%) Medicare 20.6 (±0.3) 15.2 (±0.4) Medicaid 15.3 (±0.3) 10.3 (±0.4) Private (including HMO) 51.2 (±0.5) 65.8 (±0.9) Other 12.7 (±0.3) 8.7 (±0.5) Median household income quartiles for patient’s zip (%) 1–38 999 24.2 (±0.6) 19.3 (±0.7) 39 000–47 999 24.9 (±0.5) 24.0 (±0.9) 48 000–62 999 25.3 (±0.5) 26.9 (±0.7) 63 000+ 25.6 (±0.9) 29.8 (±1.5) Length of stay (days) 4.7 (±0.03) 10.0 (±0.11) Total charges (US $) 23 005 (±396) 60 024 (±1 604) Charlson–Deyo 0.3 (±0.004) 0.2 (±0.007) comorbidity score Mortality (%) 0.2 (±0.02) 0.8 (±0.08) Mean UV index (units) 4.4 (±0.03) 4.3 (±0.05) Discharge quarter (%) January–March 24.3 (±0.2) 25.2 (±0.4) April–June 25.1 (±0.2) 25.6 (±0.4) July–September 25.6 (±0.2) 24.8 (±0.4) October–December 24.9 (±0.2) 24.4 (±0.4)
Downloaded from http://ecco-jcc.oxfordjournals.org/ by guest on November 14, 2015
USA.1 This north–south gradient has also been evident in population studies in France.2 Geographical areas near the equator are exposed to more ultraviolet (UV) light than those farther away, and UV light is a major factor in the endogenous production of vitamin D. It is therefore hypothesized that the gradients seen in CD are due to vitamin D production. Both the prevalence of Crohn’s disease and disease activity have been inversely associated with vitamin D levels.3,4 In a retrospective study, patients with CD and a 25-OH vitamin D level below 20ng/ml had an increased odds of surgery (OR 1.76, 95% CI: 1.24–2.51) adjusting for age, gender and other comorbidities.5 Among those patients who had multiple measurements of vitamin D levels and normalized their levels, their odds of undergoing surgery were significantly lower (OR 0.56, 95% CI: 0.32–0.98). A small randomized controlled trial of vitamin D supplementation suggested a trend in the reduction of clinical relapse in patients supplemented with vitamin D.6 We aimed to determine if there was an association between the risk of undergoing an inpatient gastrointestinal surgery and UV light exposure among patients with Crohn’s disease using the largest database of inpatient hospitalizations in the USA.
UV Associated with Reduced Crohn’s Surgery
79
Dividing the sample into approximately equal population tertiles of UV exposure, we compared the number of hospitalizations for those without CD with the number of CD-related hospitalizations (Figure 1). This analysis shows a significant difference between these two groups, with the largest percentage of CD hospitalizations occurring in areas with the lowest UV exposure and the lowest percentage in the areas with the greatest UV exposure. Univariate analysis of the odds of inpatient surgery among those with a CD diagnosis revealed that for each extra unit of UV index exposure, there was a 10% reduction in the risk of undergoing inpatient surgery (Table 2). Other notable findings included a reduced chance of surgery for females, and a significantly increased chance of surgery with increased income status. Hospitalizations at urban institutions were also significantly more likely to include a surgery.
A.
Both African Americans and Hispanic patients appear less likely to undergo surgery. In order to control for confounding factors on the odds of inpatient surgery, a multivariate model was constructed (Table 3). Accounting for age, gender, race/ethnicity, Charlson–Deyo comorbidity index, hospital location, average household income in the patient’s zip code, and season, mean UV index remains protective for inpatient surgery (OR 0.91 per unit of UV index, 95% CI: 0.84–0.98, p = 0.01). For each 3-unit increase, this translates to a 26% reduction in the risk of inpatient surgery. In this multivariable model, discharge season did not meet statistical significance. In multivariate analysis, women, African Americans and Hispanics were all less likely to undergo surgery. In sensitivity analyses (Table 4), we found that there was no significant change in our odds ratio estimates if states with high variance in UV index or Hawaii (highest UV index and few surgeries)
United States Map with UV Index by Population Equal Tertile
< 3.78 Mean State UV Index
3.78-5.47 >5.47
B.
Comparison of Hospitalizations from Each UV Index Tertile 0.5
0.4
Proportion
Legend 0.3
No primary diagnosis of Crohn’s disease
0.2
Primary diagnosis of Crohn’s disease
0.1
0.0
Lowest Tertile
Middle Tertile
Highest Tertile
Figure 1. A map of the USA demonstrates that the highest UV exposure is located in the southwest (panel A). The proportion of hospitalizations categorized by approximately equal UV index tertile with a diagnosis of CD is statistically different from the proportion of all hospitalizations (panel B).
Downloaded from http://ecco-jcc.oxfordjournals.org/ by guest on November 14, 2015
Legend
S. M. Govani et al.
80 Table 2. Univariate odds ratios of inpatient surgery among patients hospitalized with a primary Crohn’s diagnosis. OR
95% CI
p-value
Predictor
Original Article
Increased Ultraviolet Light Exposure is Associated With Reduced Risk of Inpatient Surgery Among Patients With Crohn’s Disease Shail M. Govania,*, Peter D.R. Higginsa, Ryan W. Stidhama, Scott J. Montainb, Akbar K. Waljeea,c Downloaded from http://ecco-jcc.oxfordjournals.org/ by guest on November 14, 2015
Department of Internal Medicine, Division of Gastroenterology, University of Michigan, Ann Arbor, MI, USA b United States Army Research Institute of Environmental Medicine, Natick, MA, USA c Veterans Affairs Center for Clinical Management Research, Ann Arbor, MI
a
*Corresponding author: Shail M. Govani, 1500 E. Medical Center Drive, 3912 Taubman Center, SPC 5362, Ann Arbor, MI 48109, USA. Tel: +1-734-615-8468; Fax: +1-734-763-7834; Email: [email protected]
Abstract Background and aims: Due to the formation of strictures and fistulas, patients with Crohn’s disease (CD) frequently need surgery. Vitamin D has been found to play a role in the degree of inflammation. We aimed to study the effect of UV exposure on the need for inpatient surgery in patients with CD. Methods: The national inpatient sample, the largest database of hospitalizations from the USA, was used to conduct a retrospective study of patients hospitalized from 2004–2011 with CD by ICD9CM code 555. Surgery was characterized as any intestinal surgery or fistula repair. An average state UV exposure was calculated for each hospitalization. Multivariate logistic regression was used to calculate the effect of UV exposure on surgery accounting for important covariates. Results: There were 481 712 hospitalizations with a primary diagnosis of CD. Of these hospitalizations, 67 751 included a relevant surgical procedure code. Mean UV exposure was statistically lower in the group undergoing surgery (4.3 units versus 4.4 units, p = 0.001). The ratio of hospitalizations per UV exposure tertile for CD was statistically different compared with all hospitalizations (p 11. The NIS does not contain patient level zip code level, but does contain hospital zip code data for 68.6% of the patients. As there is no zip code data for 31.4% of the patients, we averaged the UV index values to obtain state level averages for all hospitalizations and integrated this data into the NIS in order to perform our main analysis. As a sensitivity analysis, we also performed multivariable modeling using 3-digit zip code data, excluding states with high standard deviations of UV index (California and Colorado), excluding perianal fistula repairs, and excluding Hawaii (because it has very few hospitalizations). As the NIS is a national de-identified dataset, IRB approval was not obtained.
2.2 Statistical analysis SAS 9.3 (Cary, NC, USA) was used to perform the statistical analysis. Complete case analysis was used for all testing. PROC SURVEYFREQ (Rao–Scott chi-square test) and SURVEYREG (t-test) were used to compare the baseline characteristics of patients with CD who underwent surgery versus those who did not. PROC SURVEYLOGISTIC was used to perform univariate and multivariable modeling to adjust for other important factors. Appropriate weighting was used to calculate national estimates with domain level analysis to ensure accurate standard error estimates.
3. Results During the study period, there were ~481 712 hospitalizations with a primary diagnosis of Crohn’s disease, with 14.1% of hospitalizations including an intestinal surgery or fistula repair. A comparison of patients with CD who underwent surgery versus those who did not is displayed in Table 1. Patients who underwent a surgery were more likely to be younger and male. There were was a trend in the income status of patients, with patients from the lowest income status less likely to undergo surgery. Unsurprisingly, patients who underwent surgery had longer lengths of stay and hospital costs. There was a statistically significant lower mean UV index exposure in patients who underwent surgery. Among patients who underwent surgery, the average UV index was 4.3 compared with an average UV index of 4.4 among those who did not receive an operation (p = 0.001).
Table 1. Comparison of patients with CD who underwent a surgery versus those who did not. Crohn’s disease visits not including a surgery (n = 413 961)
Crohn’s disease visit including a surgery (n = 67 751)
Age (years) 40.5 (±0.2) 41.1 (±0.2) Female (%) 57.5 (±0.3) 52.0 (±0.5) Race/Ethnicity (%) Caucasian 76.9 (±0.6) 84.1 (±0.9) African American 13.5 (±0.5) 8.7 (±0.4) Hispanic 5.8 (±0.4) 3.5 (±0.3) Other 3.7 (±0.2) 3.6 (±0.7) Payer information (%) Medicare 20.6 (±0.3) 15.2 (±0.4) Medicaid 15.3 (±0.3) 10.3 (±0.4) Private (including HMO) 51.2 (±0.5) 65.8 (±0.9) Other 12.7 (±0.3) 8.7 (±0.5) Median household income quartiles for patient’s zip (%) 1–38 999 24.2 (±0.6) 19.3 (±0.7) 39 000–47 999 24.9 (±0.5) 24.0 (±0.9) 48 000–62 999 25.3 (±0.5) 26.9 (±0.7) 63 000+ 25.6 (±0.9) 29.8 (±1.5) Length of stay (days) 4.7 (±0.03) 10.0 (±0.11) Total charges (US $) 23 005 (±396) 60 024 (±1 604) Charlson–Deyo 0.3 (±0.004) 0.2 (±0.007) comorbidity score Mortality (%) 0.2 (±0.02) 0.8 (±0.08) Mean UV index (units) 4.4 (±0.03) 4.3 (±0.05) Discharge quarter (%) January–March 24.3 (±0.2) 25.2 (±0.4) April–June 25.1 (±0.2) 25.6 (±0.4) July–September 25.6 (±0.2) 24.8 (±0.4) October–December 24.9 (±0.2) 24.4 (±0.4)
Downloaded from http://ecco-jcc.oxfordjournals.org/ by guest on November 14, 2015
USA.1 This north–south gradient has also been evident in population studies in France.2 Geographical areas near the equator are exposed to more ultraviolet (UV) light than those farther away, and UV light is a major factor in the endogenous production of vitamin D. It is therefore hypothesized that the gradients seen in CD are due to vitamin D production. Both the prevalence of Crohn’s disease and disease activity have been inversely associated with vitamin D levels.3,4 In a retrospective study, patients with CD and a 25-OH vitamin D level below 20ng/ml had an increased odds of surgery (OR 1.76, 95% CI: 1.24–2.51) adjusting for age, gender and other comorbidities.5 Among those patients who had multiple measurements of vitamin D levels and normalized their levels, their odds of undergoing surgery were significantly lower (OR 0.56, 95% CI: 0.32–0.98). A small randomized controlled trial of vitamin D supplementation suggested a trend in the reduction of clinical relapse in patients supplemented with vitamin D.6 We aimed to determine if there was an association between the risk of undergoing an inpatient gastrointestinal surgery and UV light exposure among patients with Crohn’s disease using the largest database of inpatient hospitalizations in the USA.
UV Associated with Reduced Crohn’s Surgery
79
Dividing the sample into approximately equal population tertiles of UV exposure, we compared the number of hospitalizations for those without CD with the number of CD-related hospitalizations (Figure 1). This analysis shows a significant difference between these two groups, with the largest percentage of CD hospitalizations occurring in areas with the lowest UV exposure and the lowest percentage in the areas with the greatest UV exposure. Univariate analysis of the odds of inpatient surgery among those with a CD diagnosis revealed that for each extra unit of UV index exposure, there was a 10% reduction in the risk of undergoing inpatient surgery (Table 2). Other notable findings included a reduced chance of surgery for females, and a significantly increased chance of surgery with increased income status. Hospitalizations at urban institutions were also significantly more likely to include a surgery.
A.
Both African Americans and Hispanic patients appear less likely to undergo surgery. In order to control for confounding factors on the odds of inpatient surgery, a multivariate model was constructed (Table 3). Accounting for age, gender, race/ethnicity, Charlson–Deyo comorbidity index, hospital location, average household income in the patient’s zip code, and season, mean UV index remains protective for inpatient surgery (OR 0.91 per unit of UV index, 95% CI: 0.84–0.98, p = 0.01). For each 3-unit increase, this translates to a 26% reduction in the risk of inpatient surgery. In this multivariable model, discharge season did not meet statistical significance. In multivariate analysis, women, African Americans and Hispanics were all less likely to undergo surgery. In sensitivity analyses (Table 4), we found that there was no significant change in our odds ratio estimates if states with high variance in UV index or Hawaii (highest UV index and few surgeries)
United States Map with UV Index by Population Equal Tertile
< 3.78 Mean State UV Index
3.78-5.47 >5.47
B.
Comparison of Hospitalizations from Each UV Index Tertile 0.5
0.4
Proportion
Legend 0.3
No primary diagnosis of Crohn’s disease
0.2
Primary diagnosis of Crohn’s disease
0.1
0.0
Lowest Tertile
Middle Tertile
Highest Tertile
Figure 1. A map of the USA demonstrates that the highest UV exposure is located in the southwest (panel A). The proportion of hospitalizations categorized by approximately equal UV index tertile with a diagnosis of CD is statistically different from the proportion of all hospitalizations (panel B).
Downloaded from http://ecco-jcc.oxfordjournals.org/ by guest on November 14, 2015
Legend
S. M. Govani et al.
80 Table 2. Univariate odds ratios of inpatient surgery among patients hospitalized with a primary Crohn’s diagnosis. OR
95% CI
p-value
Predictor
