6.4 Data collection and tool
Data collection will be done using a paper data collection tool. It will be used to collect secondary data from the labour ward delivery registers, perinatal registers and mortality registers. The data will be collected primarily by the researcher and double entered to prevent errors. Data will also be collected from neonatal intensive care unit and special care baby unit. Hospital case notes will be retrieved and the clinical data collected.
Data Collection Sheet
Maternal characteristics
Hospital Number………………………… Date/time of entry in labour ward register.....................
Age ......./years
14-19
|
1
|
20-24
|
2
|
25-29
|
3
|
30-34
|
4
|
35-39
|
5
|
>40
|
6
|
Gravidity…………
Parity……………..
Gestational age at admission ............weeks …………days (decimals)
20-26+6
|
1
|
0.14
|
27-29+6
|
2
|
0.28
|
30-33+6
|
3
|
0.42
|
34-36+6
|
4
|
0.57
|
37-40
|
5
|
0.71
|
>40+1
|
6
|
0.86
|
Number of foetuses: Single 0 Multiple 1
Marital status: Single 0 Married 1 Divorced 2
Level of education: Nil 0 Primary 1 Secondary 2 College 3 University 4
HIV status: -ve 0 +ve 1 unknown 2
ARV’s: No 0 Yes 1
Booked status: No 0 Yes 1
Referred cases: No 0 Yes 1
Unbooked: No 0 Yes 1
Past obstetric history
Aspirin therapy: No 0 Yes 1
Past obstetric history of hypertension: No 0 Yes 1
Past medical history
Pre-existing medical conditions: No 0 Yes 1
Hypertension: No 0 Yes 1
Diabetes: No 0 Yes 1
Kidney disease: No 0 Yes 1
Area of dwelling
Place of dwelling: urban 0 rural 1
Symptoms
Symptoms: No 0 Yes 1
Nausea/vomiting: No 0 Yes 1
Frontal headaches: No 0 Yes 1
Epigastric pains No 0 Yes 1
Visual disturbances: No 0 Yes 1
Right upper quadrant pains: No 0 Yes 1
Vaginal bleeding with abdominal pains: No 0 Yes 1
Chest pains: No 0 Yes 1
Cardiovascular tests
Presenting BP
SBP…………..mmHg
<160
|
1
|
160-180
|
2
|
181-200
|
3
|
201-220
|
4
|
221-240
|
5
|
241-260
|
6
|
>260
|
7
|
DBP………....mmHg
<110
|
1
|
110-130
|
2
|
131-150
|
3
|
151-170
|
4
|
171-190
|
5
|
191-220
|
6
|
>220
|
7
|
Simple bedside renal test
Dipstick proteinuria: nil/trace 0 + 1 ++ 2 +++ 3 ++++ 4 not done 99
0
|
1
|
+
|
2
|
++
|
3
|
+++
|
4
|
++++
|
5
|
99
|
6
|
Haematological tests
Haemoglobin (Hb) ...............g/dl
0-4.99
|
1
|
5-9.99
|
2
|
10-14.99
|
3
|
>15
|
4
|
Platelet count (PLT)………....../109/l
0-49
|
1
|
50-99
|
2
|
100-149
|
3
|
>150
|
4
|
Hepatic test
Alanine amino transaminase (ALT) ...........IU/l
Therapeutic interventions
Antihypertensives: No 0 Yes 1
Magnesium sulphate: No 0 Yes 1
Corticosteroid therapy: No 0 Yes 1
Complications
Complications: No 0 Yes 1
Convulsions/CNS:No 0 Yes 1
APH: No 0 Yes 1
PPH: No 0 Yes 1
HELLP: No 0 Yes 1
Renal failure: No 0 Yes 1
CVA: No 0 Yes 1
DIC: No 0 Yes 1
Liver dysfunction: No 0 Yes 1
Liver rupture: No 0 Yes 1
ICU ventilation: No 0 Yes 1
Renal dysfunction No 0 Yes 1
Renal dialysis: No 0 Yes 1
Blood transfusion: No 0 Yes 1
FFP/Plat transfusion: No 0 Yes 1
Pulmonary oedema: No 0 Yes 1
Any other morbidity: No 0 Yes 1
Maternal death: No 0 Yes 1
Cause of death:
HELLP: No 0 Yes 1
Renal failure: No 0 Yes 1
CVA: No 0 Yes 1
DIC: No 0 Yes 1
APH: No 0 Yes 1
PPH: No 0 Yes 1
Liver rupture: No 0 Yes 1
Other cause No 0 Yes 1
Composite adverse maternal outcomes-final model
Maternal mortality or one or more serious complication of major organs morbidity in renal, hepatic, cardiac, respiratory, cerebral and haematological systems, ventilator support, pulmonary oedema, renal dialysis, transfusion of any blood product, abruption placenta and postpartum haemorrhage within 48 hours of admission to 7 days post-delivery.
Maternal death or other serious complications: No 0 Yes 1
Foetal/neonatal characteristics
Foetal heart beat present: No 0 Yes 1
Outcome Live: No 0 Yes 1
Apgar score 5 minute <7:No 0 Yes 1
Sex: Male1Female 2
Birth weight……………../g
0-500
|
1
|
501-1000
|
2
|
1001-1500
|
3
|
1501-2000
|
4
|
2001-2500
|
5
|
>2500
|
6
|
Complications: No 0 Yes 1
NICU admission: No 0 Yes 1
RDS:No 0 Yes 1
ENND: No 0 Yes 1
Cause of ENND
RDS: No 0 Yes 1
Prematurity: No 0 Yes 1
Very low birth weight: No 0 Yes 1
Sepsis: No 0 Yes 1
Congenital malformation: No 0 Yes 1
Discharged home: No 0 Yes 1
Composite adverse neonatal outcomes-final model
The composite adverse neonatal outcome will be defined as one or more of perinatal mortality, 5 minute Apgar score <7, respiratory distress syndrome and admission to neonatal intensive unit.
Candidate predictor variables for the final model development will be those variables that will be either of i) available and easy to collect in our settings including in rural health centres, ii) those that are known to be associated with preeclampsia and iii) those that are measurable, simple and reliable methods even in rural health clinics, like in the miniPIERS model by Payne et al. [30].
6.5.3 General statistical analysis
The data will be entered into a Microsoft Excel Inc. spreadsheet. Data will be exported to the SPSS Version 20 (IBM Corp., Armonk, NY, USA) for analysis. Univariate statistical analysis will be used and presented as frequencies and percentages for categorical variables. Continuous variables will be checked for normal distribution using Shapiro Wilk test and mean and standard deviation (SD) will be reported for all data. For variables not normally distributed, non-parametric tests like the Wilcoxon tests will be used. Bivariate statistical analysis will be used to test for association between independent and dependent variables, using the Pearson or Spearman two-tailed chi-square tests. This will test any statistical associations between the explanatory variables with the composite maternal and neonatal outcomes. A P value of <0.05 would be considered statistically significant.
6.6.0 Risk prediction regression model development
6.6.1 Predictor variables
Predictor variables will include the maternal characteristics, simple bedside and laboratory tests, therapeutic interventions and foetal characteristics outlined in section 6.5.2 above. Continuous variables like maternal age will be put in groups for analysis before logistic regression. Multiple imputation will be used for missing data. Multiple imputation will allow for the uncertainty about missing data, a process found in SPSS Version 20 package.
6.6.2 Composite adverse maternal and neonatal outcomes
The composite adverse maternal outcome to be predicted by the model will be determined by the Delphi consensus as described by Brown et al. and will include maternal mortality or one or more serious complication of major organ morbidity in renal, hepatic, cardiac, respiratory, cerebral and haematological systems, renal dialysis, transfusion of any blood product, abruption placenta, antepartum haemorrhage and postpartum haemorrhage within 48 hours of admission to 7 days post-delivery [39]. The composite adverse neonatal outcome will be determined by the Delphi consensus and defined as one or more of perinatal mortality, 5 minute Apgar score <7, respiratory distress syndrome and admission to neonatal intensive unit. The relationship between each predictor variable and the composite adverse maternal or neonatal outcome will first be assessed by binary logistic regression. The Hosmer-Lemeshow goodness-of- fit for logistic regression models with be used. Backward elimination regression models will be used to build models with a stopping rule of p<0.20. Predictor variables with a P value of < 0.2 will be considered for the final binary logistic regression models. Binary logistic regression models will be used to predict the adverse maternal outcome or neonatal outcome. Standard methods will be used to calculate the area under the curve (AUC) of the receiver operating characteristic (ROC) as found in SPSS Version 20.
6.6.3 The final models
In developing the final binary logistic regression models (logit), the predictor variables with a P value of < 0.2 will be considered for the following models;
where y = binary dependent variable (adverse maternal outcome or neonatal outcome)
β0 = is a constant when all variables are equated to zero
βi = is a the ith coefficient for variable i, i = 1,2,3…,k.
xi = is the ith independent variable.
6.6.4 Assessment of model’s performance and validation
Calibration ability of the model will be assessed visually by plotting deciles of predicted probability of an adverse maternal outcome against the observed rate in each decile and fitting a smooth line as done by Harrell et al., and Steyerberg et al.[40,41]. Performance of the models will be assessed using the area under the curve (AUC) of the receiver operating characteristic (ROC). Standard bootstrapping techniques will be used to assess potential over-fitting. Discrimination ability will be evaluated on the basis of area under curve of the receiver operating characteristic (AUC ROC) as stated by Hanley and McNeil [42]. Internal validation of the model will be assessed using Efron’s enhanced bootstrap method described by Efron and Tibsherani [43]. External validation will be assessed using the miniPIERS model.