Research on PGHD captured via mHealth for SSI surveillance should include the development of a database of patient-generated post-operative wound photos. Such a database would make available for research a robust data set for the examination of post-operative would health and the range of post-operative wound appearance.
PGHD captured via mHealth for SSI surveillance presents a novel opportunity to leverage the data produced for research. mHealth tools for SSI surveillance allow for the collection of post-operative wound photos taken by patients after hospital discharge.79 These data should be collected, and a database formed for their storage and use.80,81 Production of serial post-operative wound images via PGHD presents the opportunity for a large volume of data generation, and the ability to visualize and analyze wounds over time. Current practice in SSI monitoring, based in clinic/hospital settings, allows for limited data generation, and rarely the production of serial images. Serial images could impact current understandings of the progress of normal vs abnormal post-operative wound healing and provide more robust and diverse training sets for clinicians and researchers. Additional value of a wound photo database lies in the potential to utilize machine learning and computer vision methods to create algorithms for SSI surveillance, for example to determine which patients are progressing toward abnormal wound healing and may be at increased risk of SSI. Analysis of serial wound photos could also add to the knowledge base of what additional criteria may be needed to assess for SSI using mHealth tools.
In creating a database of patient-generated post-operative wound photos consider the following:
Seek opportunities for collaboration and data sharing between institutions with access to patient-generated wound photo data. Drawing on currently available data is a good starting point for the formation of a database. However, further development of the database through prospective collection of wound photo data, facilitated by current and newly deployed mHealth tools and programs for SSI surveillance, should be the eventual goal. It should also be considered whether it would be possible to tie into existing databases that house surgical data, in order to link wound photos to other clinically relevant data.
To take best advantage of the potential uses for wound photos collected from patients thousands of data points will be necessary, as well as standards for how photos are produced and analyzed. Patients take and submit photos using devices of varying quality. Device angle, lighting, and distance from the wound can all impact the quality and usability of photos taken. Adequate patient training is necessary to ensure high quality photos.
Diverse photos/samples are needed to address inherent analysis biases introduced by homogeneity of patient populations contributing to the database. It is important to capture photo data on patients who exhibit normal wound healing, as well as those who experience complications of wound healing, inclusive of those who develop SSI. Such a collection should be closely and accurately tied to reliable definitions for each of above categories. Capturing data from patients representing a range demographic backgrounds is also critical. For example, capturing wound photo data from patients across races and ethnicities is important for the development of systems to automate detection of signals of SSI across all skin types and pigmentations.