Indicator variables were used for missing values related to the intensity (20% missing values), duration (5% missing values), and age at start of smoking (3% missing values).
We further excluded 937 cases because of missing values related to gestational age.
Overall, proportions of missing values related to the quantification of medical resource use and productivity losses were small.
Proportions of missing values on the items were small with exception of missing values related to the measurement of the use of medication.
Additionally, to estimate the extent to which missing data were missing at random (that is, missing values related to observed variables), we carried out a series of sensitivity analyses using multiple imputation procedures.
In order to increase the number of samples, this study applied multiple regression imputation to fill in missing values related to monthly income, marital status and job status, which were missing in 1.6%, 0.65% and 0.51% of the cases, respectively.
To make efficient use of the data and minimize imprecise or biased estimates of associations, we imputed the missing values related to sexual risk factors by using complete cases, that is, women who completed the questionnaire.
This could be because the analysis of complete cases has limited power and may induce bias if missing values are related to covariates of interest, whereas under the missing-at-random assumption, multiple imputation corrects biases observed with complete case analysis [ 49] and increases power.
To some extent, this high rate of missing values is related to this data being obtained ten years later than the earnings data, i.e., a large share having passed away or emigrated during that time.
SQL outer joins, including left outer joins, right outer joins, and full outer joins, automatically produce Nulls as placeholders for missing values in related tables.
