Skip to content

Commit 838869e

Browse files
zmbczmbc
committed
Complete split of APH from PPH
1 parent dfabc31 commit 838869e

13 files changed

Lines changed: 652 additions & 149 deletions

File tree

‎docs/source/models/causes/maternal_disorders/gbd_2021_mncnh/antepartum_hemorrhage.rst‎

Lines changed: 380 additions & 0 deletions
Original file line numberDiff line numberDiff line change
@@ -0,0 +1,380 @@
1+
.. _2023_cause_antepartum_hemorrhage_mncnh:
2+
3+
=====================
4+
Antepartum hemorrhage
5+
=====================
6+
7+
Disease Overview
8+
----------------
9+
10+
GBD 2023 Modeling Strategy
11+
--------------------------
12+
13+
Cause Hierarchy
14+
+++++++++++++++
15+
16+
Antepartum hemorrhage does not appear in the GBD cause hierarchy.
17+
It is a *subset* of maternal hemorrhage (c_367), which is a most detailed cause in GBD 2023. The relevant portion of the GBD hierarchy is as follows:
18+
19+
- All causes (c_294) [level 0]
20+
21+
- Communicable, maternal, neonatal, and nutritional diseases (c_295)
22+
23+
- Maternal disorders and neonatal disorders (c_962)
24+
25+
- Maternal disorders (c_366)
26+
27+
- Maternal hemorrhage (c_367)
28+
29+
- Maternal hemorrhage with less than 1 liter blood loss (s_180)
30+
31+
- Maternal hemorrhage with greater than 1 liter blood loss (s_181)
32+
33+
- Mild anemia due to maternal hemorrhage (s_182)
34+
35+
- Moderate anemia due to maternal hemorrhage (s_183)
36+
37+
- Severe anemia due to maternal hemorrhage (s_184)
38+
39+
*Maternal hemorrhage (c_367)* is a most detailed cause, at level 4 of the GBD hierarchy.
40+
It has five sequelae, detailed in the following table:
41+
42+
.. list-table:: Sequelae of maternal hemorrhage
43+
:header-rows: 1
44+
:widths: 2 1 5 5
45+
46+
* - Sequela
47+
- GBD ID
48+
- Health state and disability weight
49+
- Notes
50+
* - Maternal hemorrhage with less than 1 liter blood loss
51+
- s_180
52+
- abdominopelvic problem, moderate
53+
54+
DW: 0.114 (0.078–0.159)
55+
-
56+
* - Maternal hemorrhage with greater than 1 liter blood loss
57+
- s_181
58+
- abdominopelvic problem, severe
59+
60+
DW: 0.324 (0.22–0.442)
61+
-
62+
* - Mild anaemia due to maternal haemorrhage
63+
- s_182
64+
- anaemia, mild
65+
66+
DW: 0.004 (0.001–0.008)
67+
-
68+
* - Moderate anaemia due to maternal haemorrhage
69+
- s_183
70+
- anaemia, moderate
71+
72+
DW: 0.052 (0.034–0.076)
73+
-
74+
* - Severe anaemia due to maternal haemorrhage
75+
- s_184
76+
- anaemia, severe
77+
78+
DW: 0.149 (0.101–0.209)
79+
-
80+
81+
Restrictions
82+
++++++++++++
83+
84+
The following table describes any restrictions in GBD 2023 on the
85+
effects of this cause (such as being only fatal or only nonfatal), as
86+
well as restrictions on the ages and sexes to which the cause applies.
87+
88+
.. list-table:: GBD 2023 Cause Restrictions
89+
:widths: 15 15 20
90+
:header-rows: 1
91+
92+
* - Restriction Type
93+
- Value
94+
- Notes
95+
* - Male only
96+
- False
97+
-
98+
* - Female only
99+
- True
100+
-
101+
* - YLL only
102+
- False
103+
-
104+
* - YLD only
105+
- False
106+
-
107+
* - YLL age group start
108+
- 10 to 14 (ID=7)
109+
-
110+
* - YLL age group end
111+
- 50 to 54 (ID=15)
112+
-
113+
* - YLD age group start
114+
- 10 to 14 (ID=7)
115+
-
116+
* - YLD age group end
117+
- 50 to 54 (ID=15)
118+
-
119+
120+
Vivarium Modeling Strategy
121+
--------------------------
122+
123+
Scope
124+
+++++
125+
126+
The goal of the antepartum hemorrhage model is to capture YLLs and YLDs due to
127+
antepartum hemorrhage among pregnant people.
128+
This page documents how to model the baseline burden of antepartum
129+
hemorrhage.
130+
Hemoglobin after the later ANC visit will affect the rates of antepartum hemorrhage;
131+
such effects are described on the relevant risk effects model page.
132+
133+
Summary of modeling strategy
134+
++++++++++++++++++++++++++++
135+
136+
We will not model antepartum hemorrhage as a state machine, but as a one-time decision.
137+
We will choose whether the pregnant person has antepartum hemorrhage at *some* time between the later ANC visit (if there is one) and the end of pregnancy.
138+
To obtain the decision probabilities,
139+
we will convert GBD's annual rates
140+
among females of reproductive age into conditional event rates.
141+
We will track antepartum hemorrhage
142+
deaths to calculate YLLs, and we will track incident cases by severity to calculate
143+
YLDs.
144+
145+
Cause Model Diagram
146+
+++++++++++++++++++
147+
148+
Although we're not modeling antepartum hemorrhage dynamically as a finite state
149+
machine, we can draw an analogous directed graph that can be interpreted
150+
as a (collapsed) decision tree rather than a state transition diagram.
151+
The main difference is that the values on the transition arrows
152+
represent decision probabilities rather than rates per unit time.
153+
154+
.. graphviz::
155+
156+
digraph hemorrhage_decisions {
157+
rankdir = LR;
158+
start [label="start"]
159+
end [label="end"]
160+
alive [label="parent did not die of hemorrhage"]
161+
dead [label="parent died of hemorrhage"]
162+
163+
start -> alive [label = "1 - ir"]
164+
start -> hemorrhage [label = "ir"]
165+
hemorrhage -> moderate [label = "1 - severe_fraction"]
166+
hemorrhage -> severe [label = "severe_fraction"]
167+
severe -> alive [label = "1 - cfr"]
168+
severe -> dead [label = "cfr"]
169+
moderate -> alive [label = "1"]
170+
alive -> end [label = "1"]
171+
dead -> end [label = "1"]
172+
}
173+
174+
.. list-table:: State Definitions
175+
:widths: 7 20
176+
:header-rows: 1
177+
178+
* - State
179+
- Definition
180+
* - start
181+
-
182+
* - hemorrhage
183+
- Parent simulant has antepartum hemorrhage
184+
* - moderate
185+
- Parent simulant has moderate antepartum hemorrhage (i.e., blood loss greater than 500 mL but less than 1 liter)
186+
* - severe
187+
- Parent simulant has severe antepartum hemorrhage (i.e., blood loss 1 liter or more)
188+
* - parent did not die of antepartum hemorrhage
189+
- Parent simulant did not die of antepartum hemorrhage
190+
* - parent died of antepartum hemorrhage
191+
- Parent simulant died of antepartum hemorrhage
192+
* - end
193+
-
194+
195+
.. list-table:: Transition Probability Definitions
196+
:widths: 1 5 20
197+
:header-rows: 1
198+
199+
* - Symbol
200+
- Name
201+
- Definition
202+
* - ir
203+
- incidence risk
204+
- The probability that a pregnant simulant gets antepartum hemorrhage
205+
* - severe_fraction
206+
- severe fraction
207+
- The probability that a simulant with antepartum hemorrhage has severe antepartum hemorrhage (i.e., blood loss of 1 liter or more)
208+
* - cfr
209+
- case fatality rate
210+
- The probability that a simulant with severe antepartum hemorrhage dies of that hemorrhage
211+
212+
Probabilities
213+
+++++++++++++
214+
215+
The antepartum hemorrhage cause model requires three probabilities, the
216+
incidence risk (ir) per pregnancy, the severe fraction (severe_fraction), and the case fatality rate (cfr), for use
217+
in the decision graph. The incidence risk per pregnancy will be computed as
218+
219+
.. math::
220+
221+
\text{ir} = \frac{\text{hemorrhage cases}}{\text{pregnancies}}
222+
= \frac{\text{(hemorrhage cases) / person-time}}
223+
{\text{pregnancies / person-time}}
224+
= \frac{\text{hemorrhage incidence rate}}{\text{pregnancy rate}}.
225+
226+
The severe fraction will be computed as
227+
228+
.. math::
229+
230+
\text{severe_fraction} = \frac{\text{incidence_s181}}{\text{incidence_s181} + \text{incidence_s180}}.
231+
232+
The case fatality rate will be computed as
233+
234+
.. math::
235+
236+
\begin{align*}
237+
\text{cfr} &= \frac{\text{hemorrhage deaths}}{\text{hemorrhage cases}} \\\\
238+
&= \frac{\text{(hemorrhage deaths) / person-time}}
239+
{\text{(hemorrhage cases) / person-time}}
240+
= \frac{\text{hemorrhage cause specific mortality rate}}
241+
{\text{hemorrhage incidence rate}}.
242+
\end{align*}
243+
244+
Calculating years lived with disability
245+
+++++++++++++++++++++++++++++++++++++++
246+
247+
We apply the YLDs per case for the corresponding severity level to each incident case to calculate YLDs.
248+
249+
.. math::
250+
251+
\text{ylds_per_case_severe} = \frac{\text{yld_rate_s181}}{\text{incidence_s181}}
252+
253+
.. math::
254+
255+
\text{ylds_per_case_moderate} = \frac{\text{yld_rate_s180}}{\text{incidence_s180}}
256+
257+
Note that we do *not* include YLDs for mild, moderate, or severe anemia due to antepartum hemorrhage (s_182, s_183, s_184) in our calculations because these
258+
sequelae are already counted under the anemia cause model, and we want to avoid double counting.
259+
260+
Data table
261+
++++++++++
262+
263+
The following table shows the data needed from GBD for these
264+
calculations.
265+
266+
.. note::
267+
268+
All quantities pulled from GBD in the following table are for a
269+
specific year, sex, age group, and location unless otherwise noted
270+
(e.g., SBR). Our simulation only includes pregnant women of
271+
reproductive age, so the sex will always be female. However, even
272+
though all of our simulants will be pregnant, we still pull each
273+
quantity for *all* females in a given year, age group, and location,
274+
because this is the default behavior of GBD. Since we are using the
275+
same total population in all the denominators, the person-time will
276+
cancel out in the above calculations to give us the probabilities we
277+
want.
278+
279+
.. list-table:: Data values and sources
280+
:header-rows: 1
281+
282+
* - Variable
283+
- Definition
284+
- Value or source
285+
- Note
286+
* - postpartum_fraction
287+
- fraction of maternal hemorrhage cases that are postpartum
288+
- 0.507
289+
- GBD 2023 crosswalk intercept of -0.68, exponentiated
290+
* - ir
291+
- antepartum hemorrhage incidence risk per pregnancy
292+
- (1 - postpartum_fraction) * incidence_c367 / pregnancy_rate
293+
- The value of ir is a probabiity in [0,1]. Denominator includes
294+
all pregnancies.
295+
* - incidence_c367
296+
- incidence rate of maternal hemorrhage
297+
- como
298+
- Use the :ref:`total population incidence rate <total population
299+
incidence rate>` directly from GBD and do not rescale this
300+
parameter to susceptible-population incidence rate using
301+
condition prevalence. Total population person-time is used in
302+
the denominator in order to cancel out with the person-time in
303+
the denominators of birth_rate and csmr_c367.
304+
* - incidence_s181
305+
- incidence rate of severe maternal hemorrhage
306+
- como
307+
-
308+
* - incidence_s180
309+
- incidence rate of moderate maternal hemorrhage
310+
- como
311+
-
312+
* - csmr_c367
313+
- maternal hemorrhage cause-specific mortality rate
314+
- deaths_c367 / population
315+
- Note that deaths / (average population for year) = deaths / person-time
316+
* - deaths_c367
317+
- count of deaths due to maternal hemorrhage
318+
- codcorrect
319+
-
320+
* - population
321+
- average population in a given year
322+
- get_population
323+
- Specific to age/sex/location/year demographic group. Numerically
324+
equal to person-time for the year.
325+
* - pregnancy_rate
326+
- pregnancy rate
327+
- (1 + SBR) * ASFR + incidence_c995 + incidence_c374
328+
- Units are total pregnancies per person-year
329+
* - ASFR
330+
- Age-specific fertility rate
331+
- get_covariate_estimates: coviarate_id=13
332+
- Assume lognormal distribution of uncertainty. Units in GBD are
333+
live births per person, or equivalently, per person-year.
334+
* - SBR
335+
- Stillbirth to live birth ratio
336+
- get_covariate_estimates: covariate_id=2267
337+
- Parameter is not age specific and has no draw-level uncertainty.
338+
Use mean_value as location-specific point parameter.
339+
* - incidence_c995
340+
- Incidence rate per person-year of abortion and miscarriage
341+
- como
342+
-
343+
* - incidence_c374
344+
- Incidence rate per person-year of ectopic pregnancy
345+
- como
346+
-
347+
* - yld_rate_s180
348+
- YLD rate per person-year due to moderate maternal hemorrhage
349+
- como
350+
-
351+
* - yld_rate_s181
352+
- YLD rate per person-year due to severe maternal hemorrhage
353+
- como
354+
-
355+
356+
.. todo::
357+
Get data from the GBD team to more precisely define the postpartum_fraction parameter and include uncertainty.
358+
359+
Validation Criteria
360+
+++++++++++++++++++
361+
362+
Limitations
363+
-----------
364+
365+
* Because we use the severity split and mortality rate of maternal hemorrhage overall, we are assuming that these are the same for postpartum hemorrhage as for antepartum hemorrhage.
366+
In reality, postpartum hemorrhage is likely to be more severe.
367+
* We assume that the impact of antepartum hemorrhage on hemoglobin is the same as the impact of postpartum hemorrhage on hemoglobin, which is likely not true.
368+
We suspect that APH would have a smaller impact on average; however, the MarketScan data the GBD effect is calculated from are
369+
from PPH in the US, so may already be capturing a smaller average impact than we would expect from PPH in a lower-resource setting.
370+
* We assume that all antepartum hemorrhage fatalities occur among those with severe antepartum hemorrhage, which may not be the case in reality.
371+
* We assume that postpartum hemorrhage is uncorrelated with antepartum hemorrhage, except for the causal effect through hemoglobin. In reality, there may be both confounding and a direct causal effect.
372+
* We assume that when a simulant has both antepartum hemorrhage and a later ANC visit, the ANC visit occurs before the onset of antepartum hemorrhage.
373+
Antepartum hemorrhage typically occurs late in pregnancy, so this would often be the case.
374+
Also, in real life, antepartum hemorrhage may lead directly to delivery, even if the fetus is premature,
375+
so we don't think there are likely to be many cases of ANC visits occurring after the onset of antepartum hemorrhage.
376+
* We assume that antepartum hemorrhage is uncorrelated with the length of pregnancy and the probability of stillbirth (except through hemoglobin),
377+
which is likely not true.
378+
379+
References
380+
----------

‎docs/source/models/causes/maternal_disorders/gbd_2021_mncnh/index.rst‎

Lines changed: 1 addition & 0 deletions
Original file line numberDiff line numberDiff line change
@@ -36,6 +36,7 @@ in the indicated model-building wave:
3636
.. toctree::
3737
:maxdepth: 1
3838

39+
antepartum_hemorrhage
3940
postpartum_hemorrhage
4041
maternal_sepsis
4142
obstructed_labor

0 commit comments

Comments
 (0)