Yield and Efficiency of a Population-Based Mass Tuberculosis Screening Intervention Among Persons With Diabetes in Jiangsu Province, China

Abstract

Background: The evidence-base for mass tuberculosis screening among persons with diabetes (PWD) is poor. We evaluated the yield and costs of mass screening among PWD in eastern China.

Methods: We included individuals with type 2 diabetes from 38 townships in Jiangsu Province. Screening comprised of physical examinations, symptom screening, and chest X-rays; smear and culture testing were performed through clinical triage. We assessed the yield and number needed to screen (NNS) to detect 1 tuberculosis case among all PWD, those with symptoms, and with suggestive chest X-rays. Unit costing was collected to estimate screening costs and to calculate cost per case detected. We performed a systematic review of other mass tuberculosis screening programs concentrated on PWD.

Results: Of 89 549 screened PWD, 160 were diagnosed with tuberculosis (179 cases per 100 000 persons; 95% confidence interval [CI]: 153-205). The NNS was 560 (95% CI: 513-606), 248 (95% CI: 217-279), and 36 (95% CI: 24-48) among all participants, with abnormal chest X-rays, and symptoms. The cost per case was high overall (US$13 930) but lower with symptoms (US$1037) and high fasting blood glucose levels (US$6807). From systematic review, the pooled NNS to detect one case among all PWD (regardless of symptoms or chest X-ray results) in high- versus low-burden settings was 93 (95% CI: 70-141) versus 395 (95% CI: 283-649).

Conclusions: A mass tuberculosis screening program focused on PWD was feasible however, the overall yield was low and not cost-efficient. Risk-stratified approaches may be practical among PWD in low- and medium tuberculosis burden settings.

Keywords: cost efficiency; diabetes; screening; tuberculosis; yield.

Figure 1.

City and participant-level inclusion in tuberculosis mass screening of persons with diabetes in Jiangsu Province, China.

Figure 2.

Yield and number needed to screen to detect 1 tuberculosis case of tuberculosis- and diabetes-related symptoms during a mass screening intervention among persons with diabetes in Jiangsu Province, China. Yield of tuberculosis is shown in panel A. Number needed to screen to detect 1 tuberculosis case is shown in panel B. In both plots tuberculosis-related symptoms are shown in the shaded region, whereas other symptoms are shown in non-shaded regions. Abbreviation: CI, confidence interval.

Figure 3.

The number needed to screen to detect 1 undiagnosed tuberculosis case among persons with diabetes from mass tuberculosis screening interventions found in a systematic review. The number needed to screen to detect 1 undiagnosed tuberculosis case was calculated using standard formulas described in the Methods. Settings with a high- or low-burden setting were classified as above or below 100 incident cases per 100 000 persons in the background population. We used a random effects model with DerSimonian and Laird weights to pool rates among all studies and in sub-groups. Lin et al (2012) [32] was grouped into 2 cohorts, low- and high-burden, based on descriptions in this study of differential background burden in the settings represented in the study. More details on the systematic search and descriptions of each study are listed in the Supplementary Table 7. The NNS to detect 1 undiagnosed tuberculosis case overall was 150 (95% CI: 119–205); in high- versus low-burden settings the NNS was 93 (95% CI: 70–141) versus 395 (95% CI: 283–649), respectively. Heterogeneity was high overall (I² = 95%) and in subgroup analyses for both high- (I² = 96%) and low-burden (I² = 73%) settings. Abbreviations: CI, confidence interval; NNS, number needed to screen.