Aim 3: To identify whether, sleep quality, mood or birth experience are associated with the experience of unusual thoughts during the postpartum.
8.3 Limitations of the Study 1 Sampling limitations
Convenience sampling methods were employed across both phases of the study, which may have impacted on how representative the sample was, for example only those with access to the Internet and ability to read English participated in either phase. Furthermore, demographic information indicated that participants across both phases one and two of the study were predominantly White British; married and aged 25-34. This lack of diversity limits the generalisability of the findings and it would be important for future research to include a more diverse sample of participants as amongst other factors differences in cultural contexts and beliefs are likely to play a key role in the interpretation of what are deemed to be unusual thoughts. It is also noted that across both phases of the study small participant numbers will impact on the validity and generalisability of findings. This may have been addressed in phase two through provision of paper versions of questionnaires at recruitment sites with a postal return option. Possibly increasing the size and generalisability of the sample through the potential to reach disadvantaged groups and those without Internet access.
Generalisability of the findings is also limited by the self-selecting nature of participation. Participants across both phases were those who specifically chose to opt-in to the research process, suggesting that there may be particular characteristics of this sample, which led to their interest in participation. Unfortunately, as demographic details of those who were provided with information on the study and chose not to participate were not collected, comparisons between the groups cannot be made.
It was noted that the mean self-reported HADS scores for anxiety and depression symptoms were relatively low in the current sample (HADS A = 6.42 (SD 4.20) and HADS D = 5.87 (SD =3.60). The mean anxiety score in particular was notably lower to that reported in previous non- clinical females samples (e.g. Caci, Bayle, Mattei, Dossios, Robert & Boyer (2003) HADS A= 8.57 (SD 3.56) and HADS D = 3.23 (SD 2.09)). This may be an indication that the sample of women who responded to the current survey is made up of those who experience particularly lower levels of anxiety post-birth. In the current sample this may reflect the predominance of married participants, who may feel more emotionally supported and therefore have higher levels of wellbeing. This suggests that, the findings of the current study may not generalise to women who do not share these characteristics. Interestingly, whilst both anxiety and depression symptoms are expected to be generally higher during the postpartum period, other studies have also reported similarly low levels in non-clinical postpartum samples. For example, Tuohy & McVey (2008) HADS A= 6.92 (SD 4.50) and HADS D = 5.03 (3.80) and Van Bussel, Spitz and Demyltenare (2009) mean postpartum HADS A = 4.11 (SD 2.94). The low scores across research into anxiety and depression symptoms during the postpartum period might be an indication that research samples tend to consist of women who feel able to participate due to having higher levels of wellbeing. In a replication of the study it would be important to access a more representative sample and consider whether anxiety and depression symptoms differ. In addition, the time constraints associated with a research degree impacted on the number of participants that were able to be recruited (N=60). It may be beneficial to repeat the study using a larger sample size in order to account for bias identified in statistical analysis (for example, outliers in the unusual thought data, normality assumptions), potentially strengthening the findings. In phase two it was decided not to conduct regression analyses on the data (as discussed on page 84), however further investigation and development of the findings should consider how regression analysis could be utilised to answer additional questions about predictive relationships.
8.3.2 Measurement limitations
A further consideration and possible limitation to validity is how the language used may have had bearing on the types of thoughts participants most frequently shared. For example, whilst the term ‘unusual’ was utilised in-line with psychological terminology for thoughts experienced in psychosis (e.g. unusual beliefs), it is noted that the word may have alternative interpretations and possibly carry negative connotations. The mothers in this study most frequently provided thoughts with negative content; related to harm or catastrophe. It is possible that rather than a true indication that these are thoughts most frequently experienced, higher reporting may have been an effect related to terminology. Additionally the specific nature of language used in survey item 13 (I have thought I’m not connected to my baby when not breastfeeding) may have led to under reporting of similar thoughts for mothers who were not breastfeeding. In further refinement of the questionnaire it may be beneficial to adjust the wording to ‘feeding’ or ‘when not holding or caring for my baby’ in order to increase generalisability, whilst still capturing thoughts related to disconnection from the infant.
It is also noted that some of the thought items derived from the postpartum psychosis literature were not apparent in phase one participant accounts, however were endorsed as being experienced by the survey participants in phase two. This suggests that a wider range of thoughts occur than were expressed by participants within the online focus group. It is possible that the range of thought content shared by participants within the group discussions was biased by social desirability, with participants tending not to share, or to under report, certain types of thoughts; possibly content deemed of a more sensitive nature or more commonly associated with psychotic beliefs. Women may still have been sensitive to sharing thoughts and being judged by others even within the online environment.
Whilst anonymous surveys and online data collection methods have limitations, the data collection method utilised in this research has been reported as useful in providing a safe environment for the disclosure of unusual thoughts in a non-clinical population (Freeman et al., 2005). This study has also identified and generated a number of helpful questions for use in clinical work and in future research.