Figures
Schematic illustration of the model structure. (A) Overview of the compartmental model structure, described in detail in ref. The circle denotes the interval from a patient’s first presentation to care until ultimate TB diagnosis, during which they may visit several different providers in both the public and private sectors. By improving diagnosis in the private sector, a PPIA aims to remove patients from this loop as rapidly as possible. Additionally, by providing free drugs and adherence support in the private sector, a PPIA aims to minimise the risk of long-term recurrence (bottom left compartment). (B) A simple approach for linking PPIA ‘provider coverage’ (the number of providers recruited of a given type, for example, FQ) with ‘market share’ (the proportion of provider–patient interactions captured by a PPIA). The former is relevant for costing, while the latter is relevant for the transmission model illustrated in panel A. Curves arise from the formula: Market share = (Provider coverage)k, for a given parameter k. Ideally a PPIA would first recruit the highest-caseload providers, to capture a disproportionate amount of patient–provider interactions (k=0.5, red curve). However, we also allow conservatively for less efficient provider recruitment (k=2, yellow curve). In propagating uncertainty through Bayesian melding, we allow the control parameter k to vary uniformly between 0.5 and 2. FQ, formally qualified; PPIA, Public–Private Interface Agencies; TB, tuberculosis.
Cost-effectiveness plane of PPIAs operating at scale. Different ‘dots’ of a given colour arise from uncertainty in model inputs, and in the unit costs used. Blue, red and green dots represent provider coverage scenarios of 25%, 50% and 75%, respectively. Shaded regions show willingness-to-pay thresholds listed in table 2, as follows: points in the white region are cost-effective with respect to all thresholds listed in table 2. Points in the lightest grey region are ‘highly cost-effective’ with respect to the WHO-CHOICE threshold, but not cost-effective with respect to the ‘stringent’ threshold. Points in the middle grey region are cost-effective with respect to WHO-CHOICE, and the dark grey region shows parameters that are not cost-effective under any of the thresholds listed in table 2. Both horizontal and vertical axes correspond to cumulative totals over the period 2017 to 2025.
Cost-effectiveness planes showing different prioritisations of PPIA activities. Results show the case of 50% engagement of private providers in both cities. Points in blue show a scenario where the PPIA focuses on improving diagnostic quality among private providers, without addressing patient treatment outcomes. Points in red show the converse scenario where the PPIA focuses on improving patient treatment outcomes, without addressing improved diagnostics. Points in yellow show combined efforts to improve diagnostic and treatment quality. Shaded regions show willingness-to-pay thresholds listed in table 2, as follows: points in the white region are cost-effective with respect to all thresholds listed in table 2. Points in the lightest grey region are ‘highly cost-effective’ with respect to the WHO-CHOICE threshold, but not cost-effective with respect to the ‘stringent’ threshold. Points in the middle grey region are cost-effective with respect to WHO-CHOICE, and points in the dark grey area satisfy none of the three thresholds for cost-effectiveness. In general, the greater the angle between any given point and the positive X-axis, the more favourable it is in cost-effectiveness terms. In both settings, therefore, PPIAs focusing only on treatment outcomes will be most cost-effective (red points) but in Mumbai, such a strategy would substantially compromise overall health impact, by neglecting diagnostics (comparison of red and yellow points). DALY, disability adjusted life years; PPIA, Public–Private Interface Agencies.
Illustrative impact of the different ‘arms’ of a PPIA. Shown are epidemic curves under different scenarios, assuming 50% coverage, for Mumbai (top row) and Patna (lower row). For DS-TB (left-hand panels), a PPIA concentrating on treatment has essentially the same effect as a ‘full’ PPIA that addresses both treatment and diagnosis (yellow and orange curves, overlaid). A PPIA concentrating on diagnosis, however, has little impact on incidence (blue and purple curves, overlaid). These dynamics illustrate that, for DS-TB, the incidence impact of a PPIA arises largely from controlling recurrence rates through treatment completion, rather than by reducing the delay to diagnosis. For DR-TB (right-hand panels), however, diagnosis plays a stronger role in incidence reduction than treatment: this is because of the low current rates of drug sensitivity testing at the point of TB diagnosis, that could be addressed by providing drug-susceptibility testing to patients cared for by the private sector. DS-TB, drug-susceptible tuberculosis; PPIA, Public–Private Interface Agencies.
See this image and copyright information in PMC