Paperonium

← Back to the plain-language case

Abstract's $1.3 billion TB cost total exceeds the paper's own yearly costs about tenfold

Lancet Reg Health Am (2024) Β· doi:10.1016/j.lana.2024.100905 Β· PMC11747192

Summary

The Abstract (Findings) and the Discussion give the total direct cost of pulmonary TB to Brazil's public health system over 2015–2022 as more than, or about, $1.3 billion. The Results give the same total as 108 million dollars, with yearly costs between 9.98 million (2022) and 21.38 million (2017) dollars; eight such years can add up to at most $171 million, so $1.3 billion is impossible on the paper's own numbers. The authors' public data file reproduces the yearly values exactly and sums to $131.3 million (all records) or $124.6 million (the 426,040 analysed cases); $1.3 billion is about ten times either. The figure is the abstract's headline estimate of the economic burden, and no correction has been published.

What the paper reports

Abstract, Findings:

The total direct cost for pulmonary TB in Brazil during the seven years exceeded $1.3 billion, with retreatment cases accounting for $23.5 million.

Discussion, second paragraph:

The total TB cost to the public health system over the last seven years amounted to approximately $1.3 billion. While costs varied according to each population, attention is drawn to the expenditure on retreatment cases, totalling $23.5 million.

Results, "Evaluating the direct cost of pulmonary TB from 2015 to 2022":

The overall cost of pulmonary TB treatment in Brazil exhibited variations over the years, reaching its highest value in 2017 (21.38 million dollars) and the lowest in 2022 (9.98 million dollars), accumulating to a total cost of 108 million dollars spent over the 7-year period (Fig. 2a). Further stratifying the cost based on the history of anti-TB treatment, surprisingly, despite fluctuations in the total cost and the cost of new cases, the cost of retreatment cases remained stable over the years with an accumulated total of 23,5.million dollars (Fig. 2b).

Methods, "Tuberculosis direct cost analysis", on how the total is built:

After calculating the cost per case, we summed the value of each case for each year of the study.

Results, "Characteristics of the study participants":

Following these exclusions, our study comprised a total of 426,040 pulmonary TB cases.

Results, cost needed to reach a 90% cure rate:

Initially, an investment of $ 275 was identified as necessary for the overall population to achieve a 90% cure rate. Further delineating the groups based on treatment history, the new case subgroup required a similar investment of $276 to attain the 90% cure rate, mirroring the overall population. However, the retreatment group necessitated increased efforts, with an investment of $318 for the same target (Fig. 6).

The per-group investments elsewhere in the Results run from $217 (health workers) to $366 (homeless). Table 1 has columns headed N = 345,908 (new cases) and N = 80,132 (retreatment).

What the paper's own numbers show

Method. The Methods define the total as the sum of per-case costs, added up by year. The Results give the largest and smallest yearly totals, which bound any total over the study period: 2015–2022 spans eight calendar years (the paper says seven), so the total must lie between 8 Γ— 9.98 and 8 Γ— 21.38 million dollars. I then checked the total against the paper's case counts and its per-case costs. Finally, I summed the per-case cost column (ind_cost) of the data file the paper links in its Data sharing statement (data/SINAN_TB.parquet in the authors' public repository, commit 6e74aaf of 2024-09-09, SHA-256 checked), following the authors' own code (scripts/Main_code.Rmd), which builds Fig. 2a as the sum of ind_cost by year over all records. Every number below is printed by recompute.py.txt.

Quantity Location Reported Recomputed
Total direct cost, 2015–2022 Abstract (Findings); Discussion "exceeded $1.3 billion"; "approximately $1.3 billion" Not possible: at most $171.0M from the stated yearly maximum
Total direct cost, 2015–2022 Results (Fig. 2a) 108 million dollars Within the bounds: $69.9M–$149.7M (7 years), $79.8M–$171.0M (8 years)
Ratio of the two totals Abstract vs Results $1.3B vs $108M 12.04Γ—
Years at the 2017 maximum needed to reach $1.3B Results 7-year period 60.8 years
Cost per analysed case (426,040) Results, Fig. 1 not given $253 if $108M; $3,051 if $1.3B
Paper's modelled cost per case for 90% cure Results, Figs 6–7 $217–$366 (overall $275) $1.3B per case is 8.3Γ— the highest
$1.3B spread over all 743,823 notified TB cases Results not given $1,748 per case
Retreatment ($23.5M) as share of total Abstract; Results; Table 1 $23.5M 21.8% of $108M (retreatment is 18.8% of cases); 1.8% of $1.3B
Implied cost per new case vs per retreatment case Table 1 counts retreatment needs more ($318 vs $276) $244 vs $293 if $108M; $3,690 vs $293 if $1.3B
Yearly totals, authors' data (million USD) Fig. 2a; Results 21.38 (2017), 9.98 (2022) 13.98, 16.48, 21.38, 20.81, 20.06, 14.02, 14.60, 9.99 (2015–2022)
Total, authors' data, all 458,297 records Fig. 2a 108 million dollars $131.3M
Total, authors' data, 426,040 analysed cases Results; Table 1 β€” $124.6M (row counts 345,908 new and 80,132 retreatment match Table 1)
New cases only, authors' data, all records Fig. 2b β€” $107.8M
Retreatment only, authors' data, all records Results; Fig. 2b 23,5 million dollars $23.6M
Highest cost of any single case authors' data β€” $490.24; 426,040 cases at that cost = $208.9M
$1.3B / data total β€” β€” 9.90Γ— (all records); 10.44Γ— (analysed cases)
Data total in BRL, year by year at Table S3 rates Supplementary Table S3 β€” R$530M (0.53B BRL)

Every check in the paper's own text and data points to a total of roughly $100–$130 million. The $1.3 billion in the Abstract and Discussion is about 12 times the Results figure and about 10 times what the authors' data add up to.

A second, smaller point appears in the data. The Results call 108 million dollars the overall total, but the authors' data and the yearly values plotted in Fig. 2a sum to $131.3 million. $108 million matches the new-case subtotal ($107.8 million), and $107.8M + $23.6M (retreatment) = $131.3M. The Results total therefore appears to leave out retreatment cases. This does not change the main finding: every candidate total lies between about $108 and $131 million.

Figure

The Abstract and Discussion give $1,300 million for the same total that the Results give as $108 million

Figure 1. The two statements of the same quantity, the total direct cost of pulmonary TB over 2015–2022. Left: the Abstract (Findings) and Discussion, $1,300 million. Right: the Results, which give the yearly maximum and minimum and a total of $108 million; the two differ by a factor of 12. Orange marks values printed in the paper. The blue lines below are recomputed by recompute.py.txt: the range an eight-year total can take given the stated yearly extremes, and the totals of the authors' public data.

Dot plot on a linear axis from 0 to 1,400 million US$: the Abstract value of 1,300 lies far from every other value, all of which fall between 80 and 171

Figure 2. The same totals on one linear axis that starts at zero, in million US$. Orange dots: values reported in the paper (Abstract and Discussion, 1,300; Results, 108). Blue: recomputed values (the range 80–171 allowed by the stated yearly maximum and minimum over eight years; the authors' data summed over all records, 131, over the 426,040 analysed cases, 125, and over new cases only, 108). Only the Abstract value lies outside the range the paper's own yearly figures allow.

Why it matters

Explanations considered

Status

Suggested correction

In the Abstract (Findings) and the Discussion (second paragraph), the total direct cost should be replaced by the total from the study's data. The authors should confirm which population the headline total covers. Summed over all records, as in Fig. 2a, it is about $131 million; over the 426,040 analysed cases it is about $125 million. For example: "The total direct cost for pulmonary TB in Brazil during 2015–2022 was approximately $131 million, with retreatment cases accounting for $23.5 million." The same figure should replace "approximately $1.3 billion" in the Discussion. The Results sentence giving "a total cost of 108 million dollars" should be checked as well, since $108 million matches the new-case subtotal rather than the overall total shown in Fig. 2a. "23,5.million" should read "23.5 million". The authors may also wish to say "eight years" (2015–2022) instead of "seven years" in the title, Abstract, Results and Discussion.