Estimating Partnership Duration among MSM in Belgium—A Modeling Study

Round 1

Reviewer 1 Report

Comments and Suggestions for Authors

I appreciate your paper and your efforts to describe a possible model of homophily and 

Moving from your first line in introduction, I really was expecting a speculation in your discussion regarding your model and the risk of acquire STDs. Really to speculate about the number of partner and the lenght of the relationship could be unsufficient if we don't have any information about the type of intercourse, the safe or unsafe sex, the use of condom since the beginning of the intercourse, the perception of the risk during oral sex and so one.

It could be useful in my opinion to speculate about these aspects

Comments on the Quality of English Language

with an raising concern regarding NG, 31

 as well as well as studies from Hui et al14–16 and Kasaie et al18, all reporting an 215

 

 between 4-6 and 8-13 (days) for a HA-218

 

Author Response

I appreciate your paper and your efforts to describe a possible model of homophily and moving from your first line in introduction, I really was expecting a speculation in your discussion regarding your model and the risk of acquire STDs. Really to speculate about the number of partner and the lenght of the relationship could be unsufficient if we don't have any information about the type of intercourse, the safe or unsafe sex, the use of condom since the beginning of the intercourse, the perception of the risk during oral sex and so one.

It could be useful in my opinion to speculate about these aspects

Response: We thank the reviewer for their remark. We have added a short paragraph highlighting this limitation of our study: “Although the current study tries to provide viable parameters for the network structure, it fails to link how these findings could be translated into the transmission and epidemiology of CT and NG in the MSM population. Several important aspects affecting STI transmission, such as the type and frequency of intercourse, condom use behaviors, the perception of the risk during oral sex, which are necessary for modelling STI transmission were not included in the current study. Further research is required to describe and explain how the network structure affects STI epidemiology.”

 

Comments on the Quality of English Language

with an raising concern regarding NG, 31

 as well as well as studies from Hui et al14–16 and Kasaie et al18, all reporting an 215

  between 4-6 and 8-13 (days) for a HA-218

Response: All suggested edits were amended and a thorough language check was performed.

Reviewer 2 Report

Comments and Suggestions for Authors

This is an interesting modelling study, estimating partnership duration and homophily rates among MSM in Poland. The study highlighted that they used different methodology with other studies, it should be highlighted what are the main differences in methodology compared to other study and what is the benefits of using their methodology. Overall, not sure whether their findings and methodology are interest of others.

 

The study did not specify whether Ethics has been obtained to conduct this study.

The study distinguished 3 main high activity groups, PrEP users, >15 partners, and >15 casual partners. Why >15 was selected as high activity group and reason behind?

The references in documents have errors, especially for the calibration part which is highly important.

The algorithm is iterated until the sum of errors is less than 1.5, what is the reason of choosing 1.5?

What is the novelty of using their methodology compared to others should be mentioned.

They calibrated their modelling outputs to Mendez-Lapez et al., then why not use their numbers if you calibrate to get their estimates? The modelling outputs should be validated to the empirical data to see whether their outputs are accurate enough.

In supplementary material Table S1, parameters for LA-MSM and HA-MSM among more than 15 casual partner categories are same. What was the reason?

Again, they mentioned that their methodology has many assumptions to get their modelling outputs. This exercises normally done in the modelling process, as part of estimating variable estimates, not sure whether this paper is strong enough to be published as a separate paper.

 

Author Response

This is an interesting modelling study, estimating partnership duration and homophily rates among MSM in Poland. The study highlighted that they used different methodology with other studies, it should be highlighted what are the main differences in methodology compared to other study and what is the benefits of using their methodology. Overall, not sure whether their findings and methodology are interest of others.

 

The study did not specify whether Ethics has been obtained to conduct this study.

Response: We used already collected survey data from the EMIS-2017 study. According to the data sharing agreement between the authors of this paper and the London School of Hygiene & Tropical Medicine, it is stated that the EMIS-2017 received a favorable ethical opinion from the Observational Research Ethics Committee at the London School of Hygiene and Tropical Medicine (review reference 14421 /RR/8805) on 31 July 2017. This information is included in the manuscript under Institutional Review Board Statement.

The study distinguished 3 main high activity groups, PrEP users, >15 partners, and >15 casual partners. Why >15 was selected as high activity group and reason behind?

Response: We used previously published definitions of high activity groups for comparability reasons. The definition of high-activity MSM with more than 15 partners in 12 months is mentioned in Hansson et al (2020), while the definition of more than 15 casual partners in a period of 12 months is mentioned in Cox et al (2011). We understand that such definitions could be considered arbitrary and counterarguments could be made towards using different cut-offs, but our purpose was to be able to compare our findings to other previously published papers.

The references in documents have errors, especially for the calibration part which is highly important.

Response: We apologize for this oversight. The error messages were internal references of Tables 1 and 2 in the manuscript. We manually corrected these issues. We reviewed again our list of referenced sources and we could not find an error.

The algorithm is iterated until the sum of errors is less than 1.5, what is the reason of choosing 1.5?

Response: The algorithm for estimating the counts of partners from a categorical variable was computationally-expensive.  We used 1.5 as an acceptable cutoff for the sum of errors in each of the 5 categories ("11-20", "21-30", "31-40", "41-50", "More than 50") with mean number of partners ranging between 16.4 and 128.2. Setting lower tolerances could cause problems of non-convergence, or would require very long computation times, and were decided against as they would not significantly improve the results. 

What is the novelty of using their methodology compared to others should be mentioned.

Response: Network-based models, differ from individual-based models in the sense the simulated epidemics are a result of the network structure and not of individual behaviors. This means that differently configured networks could produce substantially different behaviors in terms of epidemics. In this sense, one first needs to estimate the network structure and then simulate the epidemic events of interest on the estimated network. Although, Approximate Bayesian Computation and other methods are regularly used to calibrate parameters in order to match certain epidemiological statistics, to our knowledge, this step mainly takes place after the estimated network has been finalized. In our paper, we suggest using this commonly used calibration technique to estimate crucial and often overlooked parameters to produce a network that could replicate observed behavioral characteristics. Given the scarcity of social parameters like the ones estimated in our paper, we believe that these results are substantial enough, to be published by themselves and not just as a part of the calibration process of a bigger modeling paper. Throughout the paper, we have mentioned the lack of reliable information on partnership duration and homophily rates in the literature and have joined the call for designing and conducting studies to inform parameters of transmission models and prove or disprove the estimated parameters from our study. This lack of information could be one reason for the small number of individual- and network-based modeling studies in Europe. By utilizing the suggested methodology, we might be able to diminish one of the obstacles for more complex modeling studies in European populations. We have added a paragraph in the discussion reading: “Although the proposed methodology has been regularly used in the calibration phase of modeling studies, to the best of our knowledge, it has not been used for estimating the underlying network.”.

They calibrated their modelling outputs to Mendez-Lapez et al., then why not use their numbers if you calibrate to get their estimates? The modelling outputs should be validated to the empirical data to see whether their outputs are accurate enough.

Response: The cumulative numbers of partners reported in the EMIS-2017 study were categorical variables, with categories “more than 10” for steady partners and “11-20”, “21-30”, “31-40”, “41-50” and “more than 50” for non-steady partners. By using the estimates provided by Mendez-Lopez et al, we calculated the number of steady and non-steady partners as counts, which were then used as target statistics. The location and spread statistics between the observed (Mendez-Lopez) and calculated numbers of partners were very similar. Equally interesting was the shape of the distribution, which could not be inferred from either the primary EMIS-2017 data or the published Mendez-Lopez estimates.  We added a validation of the simulated results using the actual estimates from EMIS-2017 data in the supplementary material, to compare the distribution of observed and simulated numbers of partners.

In supplementary material Table S1, parameters for LA-MSM and HA-MSM among more than 15 casual partner categories are same. What was the reason?

Response: We thank the reviewer for indicating this mistake from our part. The correct numbers in the low-activity group have now been amended in the supplementary material.

Again, they mentioned that their methodology has many assumptions to get their modelling outputs. This exercises normally done in the modelling process, as part of estimating variable estimates, not sure whether this paper is strong enough to be published as a separate paper.

Response: We would like to refer the reviewer to our response on the novelty of our study, as a response to this remark.

Reviewer 3 Report

Comments and Suggestions for Authors

The authors conducted a study estimating partnership duration among MSM in Belgium using a network modeling approach. Overall the study is well-written and informative; I primarily have recommendations related to clarity:

 

1. Some examples of the dating apps and social media sites used to recruit participants would be helpful. These can be added in parentheses.

2. Additional description of the Separable Temporal Exponential-family Random Graph Model approach will be very helpful.

3. Specifying criteria for PrEP eligibility in the Methods will be helpful as well.

4. Some results have "Error! Reference source not found" in the text, which needs to be corrected.

5. How could social desirability bias (e.g., underreporting of casual partners) affect models. This should be described in detail.

6. Additional elaboration on research and policy recommendations will be helpful.

Author Response

Please see the attachment

Author Response File: Author Response.pdf

Round 2

Reviewer 3 Report

Comments and Suggestions for Authors

The authors have addressed my concerns.