The Missing Midwife: The Worker Bangladesh Did Not Hire

Executive Summary. Bangladesh's health workforce is missing one particular occupation, and the shortfall shows up exactly where that occupation works. In 2019, 59% of Bangladeshi births had a skilled attendant, the lowest of the four countries reporting, against a peer median of 94.7%. Two in five births, 41%, had none. The shortage is not general. Bangladesh had 0.677 physicians per 1,000 people in 2021, fourth of the five countries reporting and 75.4% of the peer median, an unremarkable deficit. Nurses and midwives ran at 0.477 per 1,000 in 2019, where the peer median of 3.069 is 6.43 times as dense. The gap against peers is wider the closer the measure sits to the delivery itself. Neonatal mortality was 17.9 per 1,000 live births in 2024, second highest of the six and 1.95 times the peer median of 9.2, while maternal mortality, at 115, ranks third of six. And 79.31% of health spending is paid out of pocket, the highest of the six, so the attendant a household does find is one it pays for in cash.

Two in five births

In 2019, skilled health staff attended 59% of births in Bangladesh. Thailand 99.1%, Indonesia 94.7%, Pakistan 73.7%. Bangladesh has the lowest share of births with skilled attendance of the four countries reporting that year. India and Vietnam have no comparable 2019 observation, so this is a comparison of four, not six. Against the peer median of 94.7%, Bangladesh is 35.7 percentage points short and stands at 62.3% of the median.

Source: World Bank World Development Indicators, births attended by skilled health staff (SH.STA.BRTC.ZS), 2019.

Two qualifications belong here rather than in a footnote. First, 2019 is the latest year the World Bank series carries for all four of these countries, so the reading is pre-pandemic and this essay cannot say where the number sits now. Second, "skilled attendant" is a definitional category, and countries differ in which cadres they count. A comparison of four national statistical conventions is not a comparison of four identical things.

What survives both caveats is the size of the gap. A 35.7 point shortfall against the median is not a measurement quibble. It is roughly the distance between attending almost every birth and attending three in five.

The gap is nursing, not medicine

The usual diagnosis of a poor country's health workforce is that it lacks doctors. That is not what this data says.

Bangladesh had 0.677 physicians per 1,000 people in 2021, fourth of the five countries reporting. Pakistan 1.16, Vietnam 1.107, Indonesia 0.688, Thailand 0.541. Bangladesh is below the peer median of 0.8975 by 0.22 per 1,000, which is to say it holds 75.4% of the median comparator's physician density. That is a gap, but it is the kind of gap a middle-income health system routinely runs.

Now the nursing side. Nurses and midwives in Bangladesh stood at 0.477 per 1,000 people in 2019, third of the four countries reporting, above only Pakistan's 0.453. India and Vietnam do not report the series for that year. The rank flatters the picture, because the two comparators above Bangladesh are far above it: Indonesia 3.781, Thailand 3.069. The peer median of 3.069 is 6.43 times the Bangladeshi density, a gap of 2.592 nurses and midwives per 1,000 people. Bangladesh runs at 15.5% of the median comparator's nursing and midwifery workforce.

Source: World Bank WDI, nurses and midwives per 1,000 people (SH.MED.NUMW.P3), 2019.

Set the two densities side by side and the structure of the workforce becomes visible. Bangladesh reports 0.677 physicians and 0.477 nurses and midwives per 1,000 people, readings two years apart, with the doctors ahead. Indonesia reports 0.688 physicians against 3.781 nurses and midwives. Thailand reports 0.541 physicians against 3.069. In both of those systems the nursing cadre is several times the medical one. In Bangladesh the ordering reverses, and among the countries reporting both series only Pakistan shares that inversion, with 1.16 physicians against 0.453 nurses and midwives.

This matters for births specifically. A delivery is not a diagnostic problem that a scarce specialist solves once. It is a stretch of hours during which someone competent has to be present, in a place the mother can reach, starting at an hour nobody scheduled. That is a job description for a cadre deployed in numbers, not for one deployed in scarcity.

The physical plant points the same way. Bangladesh had 0.94 hospital beds per 1,000 people in 2021, the lowest of the four countries reporting (Pakistan and Vietnam are absent from that year), against a peer median of 1.59. That is 0.65 beds per 1,000 below the median, or 59.1% of it. There are neither the people nor the places.

Where a missing attendant would show up

If the birth attendance gap has a cost, the cost should appear in the deaths that occur around delivery rather than in mortality generally. The two mortality series here are consistent with that, without proving it.

Neonatal mortality, deaths in the first month of life, was 17.9 per 1,000 live births in 2024, second highest of the six countries. Only Pakistan, at 36.1, is higher. India 16.7, Indonesia 9.2, Vietnam 8.8, Thailand 5.1. Bangladesh sits 8.7 above the peer median of 9.2 and carries 1.95 times the median rate.

Maternal mortality is the softer of the two readings. At 115 deaths per 100,000 live births in 2023, Bangladesh ranks third of six, above the peer median by 35 and at 1.44 times it, but below Pakistan's 155 and Indonesia's 140. India, at 80, sets the median; Vietnam is 48 and Thailand 34. Bangladesh is mid-table on maternal deaths and near the top of the group on neonatal deaths, and the essay should not smooth that difference away.

Three confounders before anyone reads causation into the pairing. Maternal mortality in this series is a modelled estimate, not a count, and the model draws on the same survey infrastructure that reports attendance, which means the two variables are not independently measured. Death registration completeness differs across six countries, and a country that records more neonatal deaths will score worse than one that records fewer, irrespective of how many occur. And attendance, workforce, and outcomes here come from four different years: 2019 for attendance and nurses, 2021 for physicians and beds, 2023 for maternal deaths, 2024 for neonatal deaths. Six countries observed at different moments is a comparison, not an identification strategy, and nothing in this data establishes that the missing attendant caused the death.

What the country already proved

The reason to take the workforce argument seriously is that Bangladesh has already done the hard part once.

Maternal mortality fell from 870 per 100,000 live births in 1990 to 115 in 2023, a fall of 755, or 86.78%. The 2023 level is 13.22% of the 1990 level. Neonatal mortality fell from 65.7 per 1,000 live births in 1990 to 17.9 in 2024, a fall of 47.8, or 72.75%.

Source: World Bank WDI, maternal mortality ratio per 100,000 live births (SH.STA.MMRT), Bangladesh against comparators, 1990-2023.

Those declines happened while skilled attendance was still low, which is the point. Falls of that size from a base of 870 do not require a midwife at every birth, because the largest early gains come from things that reach women outside the delivery room: fewer very high-parity births, tetanus immunisation, antenatal contact, better roads to a facility when something goes wrong. Bangladesh bought a great deal of survival that way. What the trend does not show is the gap closing evenly. The residual distance from this group is wider on the neonatal side, 1.95 times the peer median against 1.44 on the maternal ratio, which is what you would see if the deaths still occurring were concentrated in the hours where population-level measures stop helping and a trained person in the room starts to.

The bill lands at the same moment

Out-of-pocket payments were 79.31% of Bangladeshi health spending in 2023, the highest of the six countries and 2.02 times the peer median of 39.2%. Thailand 9.93%, Indonesia 31.08%, Vietnam 39.2%, India 43.89%, Pakistan 52.93%. Bangladesh is 40.1 percentage points above the median.

This is the financing detail that makes a workforce shortage bite twice. When skilled attendance is scarce, the households that obtain it are the ones that can pay for it at the point of delivery, in cash, at a moment chosen by the pregnancy rather than by the household budget. A scarce service financed out of pocket is not rationed by need. The data does not break out what a delivery costs, and this essay does not claim to know; what it shows is the financing regime the scarcity sits inside.

What would change this conclusion

A specific, checkable test. If, by the 2030 WDI vintage, skilled attendance in Bangladesh has climbed toward the current peer median of 94.7% while nurses and midwives remain near 0.477 per 1,000, the argument here is wrong: attendance would have been produced without the cadre this essay treats as the constraint, presumably by physicians, community workers or the private clinic sector. If attendance is still nearer 59% than 94.7% and the nursing density has not moved, the constraint will have held for a decade in plain view.

Three moves, with owners and signals.

1. Recruit and deploy against a nursing density target, not a facility count. The binding number is staff per 1,000 population, and Bangladesh's 0.477 is the figure to move. Owner: Directorate General of Nursing and Midwifery, with the Health Services Division on posting. Success signal: the WDI nurses and midwives series clearing the physician density of 0.677 per 1,000, which would give Bangladesh the staffing order Indonesia and Thailand already run and would still leave it far below the peer median of 3.069.
2. Post attendants where the unattended births are, and publish the deployment map. A national density figure can rise while the 41% of births with no skilled attendant stays where it is, if new posts land in the places that already have staff. Owner: Health Services Division. Success signal: skilled attendance above Pakistan's 73.7% in the next comparable WDI vintage, which would still leave Bangladesh below the peer median of 94.7% but would be the first movement in the series since 2019.
3. Measure attendance annually and publish it. The most recent comparable reading in this data is 2019. A country cannot manage a workforce gap it observes once every several years. Owner: Bangladesh Bureau of Statistics, alongside its regular survey rounds. Success signal: an annual published attendance series with district detail, which would make the deployment question answerable without waiting for the next international vintage.

The counterargument

The strongest objection is empirical and it is a good one. Bangladesh cut maternal deaths by 86.78% and neonatal deaths by 72.75% since 1990 while running one of the thinnest nursing workforces among the four countries reporting. If the country achieved that with 0.477 nurses and midwives per 1,000, the workforce is evidently not the binding constraint, and the money would do more elsewhere.

Three answers, one of which is a concession.

First, the objection reads a level off a trend. Falls of that magnitude come off a base of 870 maternal deaths per 100,000 live births, and the cheap interventions that produced them do not repeat. What worked to get from 870 to 115 is not what gets from 115 to Thailand's 34.

Second, the two outcome series sit at different distances from the group, in the direction the workforce argument would predict. Bangladesh is third of six on maternal deaths and second of six on neonatal deaths, and neonatal deaths are the ones most tightly bound to what happens in the hours around delivery. That is consistent with a delivery-room constraint. It is not proof of one, and a panel of six countries cannot make it proof.

Third, the concession. The attendance and nursing readings are 2019 and the outcome readings are 2023 and 2024, so the comparison pairs a pre-pandemic workforce with post-pandemic mortality, and Bangladesh's own nursing series does not stop in 2019. If the workforce has grown materially since, the gap described here is a description of the recent past rather than the present, and this data cannot tell the difference. Nor can it separate the workforce from everything correlated with it: household income, female education, urbanisation, road access, and the growth of private facility deliveries all plausibly move with skilled attendance, and none of them are in these seven indicators. The honest claim is narrow. Bangladesh, a country of 173.6 million people with a US$450 billion economy growing at 4.2% in 2024, attends a smaller share of its births with skilled staff than any of the three peers reporting alongside it, and staffs the cadre that attends births at 15.5% of the peer median. Whatever else is true, those two facts describe a choice about who gets hired.

---

Data sources: World Bank World Development Indicators, retrieved from the BDPolicyLab data lake, 2026-08-10. Comparators are India, Pakistan, Indonesia, Thailand and Vietnam; "peer median" is the median of the comparators reporting in the stated year, excluding Bangladesh. Years differ by indicator because coverage does: skilled birth attendance and nurses and midwives are 2019, physicians and hospital beds 2021, maternal mortality and out-of-pocket spending 2023, neonatal mortality 2024. Skilled attendance and the nursing workforce report for four countries in 2019 (India and Vietnam absent), physicians for five in 2021 (India absent), and hospital beds for four in 2021 (Pakistan and Vietnam absent); every comparison drawn from them is stated on that basis. Maternal mortality is a modelled estimate rather than a registered count.

Sources

1. World Bank WDI, Births attended by skilled health staff (% of total), SH.STA.BRTC.ZS: https://data.worldbank.org/indicator/SH.STA.BRTC.ZS
2. World Bank WDI, Nurses and midwives (per 1,000 people), SH.MED.NUMW.P3: https://data.worldbank.org/indicator/SH.MED.NUMW.P3
3. World Bank WDI, Physicians (per 1,000 people), SH.MED.PHYS.ZS: https://data.worldbank.org/indicator/SH.MED.PHYS.ZS
4. World Bank WDI, Hospital beds (per 1,000 people), SH.MED.BEDS.ZS: https://data.worldbank.org/indicator/SH.MED.BEDS.ZS
5. World Bank WDI, Maternal mortality ratio (per 100,000 live births), SH.STA.MMRT: https://data.worldbank.org/indicator/SH.STA.MMRT
6. World Bank WDI, Mortality rate, neonatal (per 1,000 live births), SH.DYN.NMRT: https://data.worldbank.org/indicator/SH.DYN.NMRT
7. World Bank WDI, Out-of-pocket expenditure (% of current health expenditure), SH.XPD.OOPC.CH.ZS: https://data.worldbank.org/indicator/SH.XPD.OOPC.CH.ZS